docs: point the /admin migration row at spinifex#1345

docs: record the /admin move into the spinifex registration
refactor: serve /admin through the spinifex registration
2026-10-30 16:06:22 +00:00 · 2026-10-30 10:50:52 +10:00 · 2026-10-30 10:24:04 +10:00 · 2026-10-30 10:23:37 +10:00 · 2026-10-30 10:07:21 +10:00 · 2026-10-30 09:59:15 +10:00
355 changed files with 14757 additions and 7152 deletions
@@ -223,5 +223,27 @@
    "runner": "ci-single",
    "extra": "nat-single",
    "suite": ""
+  },
+  {
+    "cell": 37,
+    "install": "binary",
+    "topo": "real-multi",
+    "net": "dhcp",
+    "host": "oracle-10",
+    "arch": "x86_64",
+    "runner": "ci-multi",
+    "extra": "",
+    "suite": ""
+  },
+  {
+    "cell": 38,
+    "install": "binary",
+    "topo": "real-multi",
+    "net": "nat",
+    "host": "oracle-10",
+    "arch": "x86_64",
+    "runner": "ci-multi",
+    "extra": "nat-multi",
+    "suite": ""
  }
 ]
@@ -102,9 +102,9 @@ jobs:
        env:
          GH_TOKEN: ${{ github.token }}

-      - name: Build microVM artifacts (lb-agent + kernel + initramfs + qboot)
+      - name: Build microVM artifacts (lb-agent + kernel + initramfs + qboot + SQS store image)
        if: steps.version.outputs.exists != 'true'
-        run: cd spinifex && make build-microvm-image
+        run: cd spinifex && make build-microvm-image build-sqs-store-microvm-image

      - name: Build amd64 Tarball
        if: steps.version.outputs.exists != 'true'
@@ -10,7 +10,7 @@ on:
  workflow_dispatch:
    inputs:
      cells:
-        description: "Comma-separated cell numbers to run, no spaces (blank = all). 1-3, 8-10, 17-19 tofu permutation matrix, 34-36 Oracle Linux single node (basic, reboot, routed NAT), 20 baremetal, 21-22 ISO, 23 predastore throughput, 24 predastore fault injection, 25-27 service suites (eks, ecs, rds), 28 guest disk performance, 29 storage fault injection, 30 routed NAT on three nodes, 31 instance recovery on three nodes, 32 load balancer recovery on three nodes, 33 KV bucket quorum under node loss on three nodes"
+        description: "Comma-separated cell numbers to run, no spaces (blank = all). 1-3, 8-10, 17-19 tofu permutation matrix, 34-36 Oracle Linux single node (basic, reboot, routed NAT), 37-38 Oracle Linux on three nodes (multi-node subset with IPsec, routed NAT), 20 baremetal, 21-22 ISO, 23 predastore throughput, 24 predastore fault injection, 25-27 service suites (eks, ecs, rds), 28 guest disk performance, 29 storage fault injection, 30 routed NAT on three nodes, 31 instance recovery on three nodes, 32 load balancer recovery on three nodes, 33 KV bucket quorum under node loss on three nodes"
        required: false

 permissions:
@@ -536,6 +536,13 @@ jobs:
          log-dir: ${{ env.ARTIFACT_DIR }}
          title: "E2E failure analysis"

+      # The WAN capture is most of the artifact and only matters when a run
+      # failed. Anything short of a clean pass (suite rc, a failed setup or
+      # collect step, a missing rc) keeps it.
+      - name: Drop the packet capture from a passing single-node run
+        if: always() && steps.e2e.outputs.rc == '0' && job.status == 'success'
+        run: rm -f "${ARTIFACT_DIR}"/net-diag/wan.pcap
+
      - uses: actions/upload-artifact@v7.0.2
        if: always()
        with:
@@ -990,6 +997,13 @@ jobs:
          log-dir: ${{ env.ARTIFACT_DIR }}
          title: "Multi-Node E2E failure analysis"

+      # The WAN captures are 90% of the artifact and only matter when a run
+      # failed. Anything short of a clean pass (suite rc, a failed setup or
+      # collect step, a missing rc) keeps them.
+      - name: Drop the packet captures from a passing multi-node run
+        if: always() && steps.e2e.outputs.rc == '0' && job.status == 'success'
+        run: rm -f "${ARTIFACT_DIR}"/net-diag-node*/wan.pcap
+
      - uses: actions/upload-artifact@v7.0.2
        if: always()
        with:
@@ -57,8 +57,8 @@ jobs:
          go-version-file: spinifex/go.mod
          cache: false

-      - name: Build microVM artifacts (lb-agent + kernel + initramfs + qboot)
-        run: cd spinifex && make build-microvm-image
+      - name: Build microVM artifacts (lb-agent + kernel + initramfs + qboot + SQS store image)
+        run: cd spinifex && make build-microvm-image build-sqs-store-microvm-image

      - name: Build amd64 Tarball
        env:
@@ -119,9 +119,6 @@ jobs:
      - name: Build Spinifex
        run: make -C spinifex build

-      - name: Manifest check + drift lint
-        run: make -C spinifex manifest-check manifest-lint
-
      - name: Run Tests with Coverage
        env:
          SPINIFEX_REQUIRE_CONFORMANCE_TOOLS: "1"
@@ -21,6 +21,9 @@ build/microvm/vmlinuz
 build/microvm/initramfs.cpio.gz
 build/microvm/qboot.rom
 build/microvm/qboot.LICENSE
+build/microvm/sqs-store/vmlinuz
+build/microvm/sqs-store/initramfs.cpio.gz
+build/microvm/sqs-store/.stage.*/
 build/qboot/
 !spinifex/runtime/roles/spinifexui/frontend/dist/
 tmp/
@@ -140,6 +140,23 @@ $(MICROVM_ARTIFACTS) &: $(MICROVM_INPUTS)
 bin/lb-agent:
 	$(MAKE) build-lb-agent

+# The SQS store microVM image: a separate artifact from the LB image above, not
+# installed or released by any target yet.
+SQS_STORE_MICROVM_OUT_DIR := $(MICROVM_OUT_DIR)/sqs-store
+SQS_STORE_MICROVM_ARTIFACTS := $(SQS_STORE_MICROVM_OUT_DIR)/vmlinuz $(SQS_STORE_MICROVM_OUT_DIR)/initramfs.cpio.gz
+SQS_STORE_MICROVM_INPUTS := scripts/build-sqs-store-microvm-image.sh scripts/lib/sqs-store-microvm-publish.sh $(SQS_STORE_MICROVM_OUT_DIR)/init.sh $(SQS_STORE_MICROVM_OUT_DIR)/inittab $(addprefix $(SQS_STORE_MICROVM_OUT_DIR)/,sqs-store-lib.sh sqs-store-prepare sqs-store-run sqs-store-stop) bin/nats-server
+
+$(SQS_STORE_MICROVM_ARTIFACTS) &: $(SQS_STORE_MICROVM_INPUTS)
+	./scripts/build-sqs-store-microvm-image.sh
+
+# The dedicated SQS store server, built from the go.mod-pinned module so the
+# image ships exactly the reviewed version. Static for the musl guest.
+bin/nats-server: go.mod go.sum
+	@mkdir -p bin
+	GOWORK=off CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -trimpath -ldflags '-s -w' -o $@ github.com/nats-io/nats-server/v2
+
+build-sqs-store-microvm-image: $(SQS_STORE_MICROVM_ARTIFACTS) ## Build the SQS store microVM kernel and initramfs
+
 # qboot is the microvm firmware. The source tarball ships with each release
 # because qboot is GPL-2.0.
 QBOOT_COMMIT := $(shell sed -n 's/^QBOOT_COMMIT="\(.*\)"$$/\1/p' scripts/build-qboot.sh)
@@ -151,6 +168,12 @@ $(QBOOT_ARTIFACTS) &: scripts/build-qboot.sh

 build-microvm-image: $(MICROVM_ARTIFACTS) $(QBOOT_ARTIFACTS) ## Build microVM kernel, initramfs and qboot firmware (incremental — skips when up to date)

+install-sqs-store-microvm: $(SQS_STORE_MICROVM_ARTIFACTS) ## Install the SQS store microVM image to /usr/share/spinifex/microvm/sqs-store/
+	sudo install -d /usr/share/spinifex/microvm/sqs-store
+	sudo rm -f /usr/share/spinifex/microvm/sqs-store/initramfs.cpio.gz
+	sudo install -m 0644 $(SQS_STORE_MICROVM_OUT_DIR)/vmlinuz /usr/share/spinifex/microvm/sqs-store/vmlinuz
+	sudo install -m 0644 $(SQS_STORE_MICROVM_OUT_DIR)/initramfs.cpio.gz /usr/share/spinifex/microvm/sqs-store/initramfs.cpio.gz
+
 install-microvm: $(MICROVM_ARTIFACTS) $(QBOOT_ARTIFACTS) ## Install microVM artifacts to /usr/share/spinifex/microvm/
 	sudo install -d /usr/share/spinifex/microvm
 	sudo install -m 0644 $(MICROVM_OUT_DIR)/vmlinuz /usr/share/spinifex/microvm/vmlinuz
@@ -158,15 +181,14 @@ install-microvm: $(MICROVM_ARTIFACTS) $(QBOOT_ARTIFACTS) ## Install microVM arti
 	sudo install -m 0644 $(MICROVM_OUT_DIR)/qboot.rom /usr/share/spinifex/microvm/qboot.rom
 	sudo install -m 0644 $(MICROVM_OUT_DIR)/qboot.LICENSE /usr/share/spinifex/microvm/qboot.LICENSE

-# the pre-commit gate: manifest checks, lint, vuln, and the unit and e2e-harness tiers.
+# the pre-commit gate: lint, vuln, and the unit and e2e-harness tiers.
 # integration and race tests skipped to keep quick, they run in CI.
 #
 # -j2 overlaps the only two heavy gates, lint and test-cover: alone each leaves
 # most of the machine idle, so pairing them cuts a cold run by about a quarter.
 # -Otarget keeps each gate's output in one block instead of interleaving them.
-# manifest-check manifest-lint test-package-check removed, can add back later
 preflight:
-	@$(MAKE) --no-print-directory -j2 -Otarget QUIET=1 lint govulncheck test-cover diff-coverage test-harness test-build-scripts manifest-check manifest-lint
+	@$(MAKE) --no-print-directory -j2 -Otarget QUIET=1 lint govulncheck test-cover diff-coverage test-harness test-build-scripts
 	@echo -e "\n ✅ Preflight passed — safe to commit."

 # Shell suites + shellcheck for build/scripts/, the systemd-unit helpers that
@@ -178,7 +200,7 @@ preflight:
 # reason: between them they install the software, wire OVN, erase the state and
 # remove the install, so a shellcheck warning in one is an outage or a wrong
 # deletion, not a style note.
-NODE_SCRIPTS := scripts/setup.sh scripts/setup-ovn.sh scripts/install-node.sh scripts/smoke-test.sh scripts/node-reset.sh scripts/uninstall-spx.sh scripts/lib/distro.sh
+NODE_SCRIPTS := scripts/setup.sh scripts/setup-ovn.sh scripts/install-node.sh scripts/smoke-test.sh scripts/node-reset.sh scripts/uninstall-spx.sh scripts/lib/distro.sh scripts/lib/stage-sqs-store-microvm.sh

 test-build-scripts:
 	@echo -e "\n....Running build/scripts/**/*_test.sh...."
@@ -186,13 +208,14 @@ test-build-scripts:
 		echo "-- $$t"; \
 		sh "$$t" || exit 1; \
 	done
-	@echo -e "\n....Running scripts/*_test.sh and scripts/lib/*_test.sh...."
-	@for t in $$( (find scripts -maxdepth 1 -name '*_test.sh'; find scripts/lib -name '*_test.sh') | sort); do \
+	@echo -e "\n....Running scripts/*_test.sh, scripts/lib/*_test.sh and scripts/appliance/*_test.sh...."
+	@for t in $$( (find scripts -maxdepth 1 -name '*_test.sh'; find scripts/lib scripts/appliance -name '*_test.sh') | sort); do \
 		echo "-- $$t"; \
 		sh "$$t" || exit 1; \
 	done
 	@echo -e "\n....Running shellcheck over build/scripts/**/*.sh + the node scripts...."
 	shellcheck -S warning $$(find build/scripts -name '*.sh' | sort) $(NODE_SCRIPTS)
+	shellcheck -S warning -x scripts/build-oracle-appliance.sh $$(find scripts/appliance -name '*.sh' | sort)
 	@echo "  test-build-scripts ok"

 # E2E harness unit tests. Build-tagged `e2e` so they're skipped by the
@@ -239,23 +262,6 @@ test-segscan-oracle:
 	@echo -e "\n....Running segscan storage oracle test...."
 	$(_Q)LOG_IGNORE=1 go test -tags=integration,segscanoracle -timeout 120s ./tests/integration/... -run TestSegscanOracle $(_RACEQ)

-# Validate docs/service-interfaces.yaml. Schema check + cross-reference
-# of services/suites/fixtures + on-disk path existence.
-manifest-check:
-	@echo -e "\n....Checking service-interfaces.yaml...."
-	@go run ./tests/e2e/manifest-check/cmd/manifest-check -repo-root . -manifest docs/service-interfaces.yaml
-
-# Drift guards: direct-create fixture lint + NATS subject lint,
-# ratcheted against tests/e2e/manifest-lint/baseline.txt. Fails only on NEW
-# drift beyond the baseline.
-manifest-lint:
-	@echo -e "\n....Linting manifest drift (fixtures + subjects)...."
-	@go run ./tests/e2e/manifest-lint/cmd/manifest-lint -repo-root .
-
-# Accept current drift into the baseline. Run after an intentional change.
-manifest-lint-update:
-	@go run ./tests/e2e/manifest-lint/cmd/manifest-lint -repo-root . -update
-
 # Run unit tests
 test:
 	@echo -e "\n....Running tests for $(GO_PROJECT_NAME)...."
@@ -468,6 +474,7 @@ distro-amd64:
 		echo "[distro-amd64] WARNING: microvm artifacts missing — tarball will not include them"; \
 		echo "[distro-amd64]          run 'make build-microvm-image' before 'make distro-amd64'"; \
 	fi
+	./scripts/lib/stage-sqs-store-microvm.sh $(SQS_STORE_MICROVM_OUT_DIR) dist/amd64
 	tar -czf dist/spinifex-$(VERSION)-linux-amd64.tar.gz -C dist/amd64 .
 	sha256sum dist/spinifex-$(VERSION)-linux-amd64.tar.gz > dist/spinifex-$(VERSION)-linux-amd64.tar.gz.sha256

@@ -486,7 +493,7 @@ distro-arm64:
 distro-clean:
 	rm -rf dist/

-.PHONY: test-package-check build build-ui build-installer build-lb-agent build-ecs-agent build-system-image build-eks-node-image import-eks-node-image publish-eks-node-image build-ecs-node-image import-ecs-node-image build-rds-postgres-image import-rds-postgres-image build-rds-mariadb-image import-rds-mariadb-image build-microvm-image install-microvm go_build preflight test test-cover test-race diff-coverage bench fuzz test-actions test-images test-build-scripts test-harness test-integration generate-aws-model-coverage aws-model-coverage test-segscan-oracle manifest-check manifest-lint manifest-lint-update \
+.PHONY: test-package-check build build-ui build-installer build-lb-agent build-ecs-agent build-system-image build-eks-node-image import-eks-node-image publish-eks-node-image build-ecs-node-image import-ecs-node-image build-rds-postgres-image import-rds-postgres-image build-rds-mariadb-image import-rds-mariadb-image build-microvm-image build-sqs-store-microvm-image install-microvm install-sqs-store-microvm go_build preflight test test-cover test-race diff-coverage bench fuzz test-actions test-images test-build-scripts test-harness test-integration generate-aws-model-coverage aws-model-coverage test-segscan-oracle \
 	deploy reinstall clean \
 	install-system install-go install-aws quickinstall \
 	lint fix govulncheck nilaway \
@@ -54,7 +54,7 @@ For Terraform, start from the [example workbooks](https://docs.mulgadc.com/docs/

 ## Services

-The platform serves **392 AWS API operations** across the services below. See exactly which ones in the [coverage report](https://docs.mulgadc.com/coverage).
+The platform serves **405 AWS API operations** across the services below. See exactly which ones in the [coverage report](https://docs.mulgadc.com/coverage).

 | Service | What it is | Built on | Status |
 | --- | --- | --- | --- |
@@ -68,8 +68,10 @@ The platform serves **392 AWS API operations** across the services below. See ex
 | **ECR** | Container registry | Predastore, deduplicated per account | Available |
 | **ECS** | Container service | Spinifex instances | Available |
 | **RDS** | Databases | PostgreSQL and MariaDB, with automated backups | Available |
+| **Route 53** | DNS | [Northstar](https://github.com/mulgadc/northstar): authoritative DNS with hosted zones | Available |
 | **Bedrock** | AI deployment | vLLM for text generation and Hugging Face TEI for embeddings and reranking | Q4 2026 |
 | **Lambda** | Serverless functions | | Q1 2027 |
+| **SQS** | Message queues | | Q1 2027 |

 Roadmap services ship under the same AWS API surface, so code written for AWS today works the moment they land. [Track what's shipped in the release notes](https://github.com/mulgadc/spinifex/releases).

@@ -29,9 +29,11 @@ var allowlist = []allowance{

 	{From: "agents/rds/agent", To: "domains/rds/instance", Debt: "agents/rds/agent (dependency, narrowed)", Reason: "command protocol, wire and record types; needs a guest/controller wire contract"},

-	{From: "cmd/aws-model-coverage", To: "domains/acm/awsapi", Debt: "cmd/aws-model-coverage", Reason: "reads declared ECR/ACM/Route 53 inventories; named exception, must not construct a registration"},
-	{From: "cmd/aws-model-coverage", To: "domains/ecr/awsapi", Debt: "cmd/aws-model-coverage", Reason: "reads declared ECR/ACM/Route 53 inventories; named exception, must not construct a registration"},
-	{From: "cmd/aws-model-coverage", To: "domains/route53/awsapi", Debt: "cmd/aws-model-coverage", Reason: "reads declared ECR/ACM/Route 53 inventories; named exception, must not construct a registration"},
+	{From: "cmd/aws-model-coverage", To: "domains/acm/awsapi", Debt: "cmd/aws-model-coverage", Reason: "reads declared service inventories; named exception, must not construct a registration"},
+	{From: "cmd/aws-model-coverage", To: "domains/ecr/awsapi", Debt: "cmd/aws-model-coverage", Reason: "reads declared service inventories; named exception, must not construct a registration"},
+	{From: "cmd/aws-model-coverage", To: "domains/route53/awsapi", Debt: "cmd/aws-model-coverage", Reason: "reads declared service inventories; named exception, must not construct a registration"},
+	{From: "cmd/aws-model-coverage", To: "domains/elbv2/awsapi", Debt: "cmd/aws-model-coverage", Reason: "reads declared service inventories; named exception, must not construct a registration"},
+	{From: "cmd/aws-model-coverage", To: "domains/rds/awsapi", Debt: "cmd/aws-model-coverage", Reason: "reads declared service inventories; named exception, must not construct a registration"},

 	{From: "runtime/daemon", To: "domains/ec2/awsapi/instance", Debt: "Instance adapter consumers outside EC2", Reason: "runtime/daemon/eks_cp_control.go and rds_deps.go call the instance adapter; needs an EC2 capability"},

@@ -124,10 +126,8 @@ var allowlist = []allowance{

 	{From: "gateway", To: "domains/eks/awsapi", Debt: "gateway (service registration)", Reason: "IAM, STS, EKS and RDS dispatch stays in gateway until the ingress registration seam"},
 	{From: "gateway", To: "domains/iam/awsapi", Debt: "gateway (service registration)", Reason: "IAM, STS, EKS and RDS dispatch stays in gateway until the ingress registration seam"},
-	{From: "gateway", To: "domains/rds/awsapi", Debt: "gateway (service registration)", Reason: "IAM, STS, EKS and RDS dispatch stays in gateway until the ingress registration seam"},
 	{From: "gateway", To: "domains/sts/awsapi", Debt: "gateway (service registration)", Reason: "IAM, STS, EKS and RDS dispatch stays in gateway until the ingress registration seam"},
 	{From: "gateway", To: "domains/ecs/awsapi", Debt: "gateway (service registration)", Reason: "ECS and ELBv2 dispatch stays in gateway until the ingress registration seam"},
-	{From: "gateway", To: "domains/elbv2/awsapi", Debt: "gateway (service registration)", Reason: "ECS and ELBv2 dispatch stays in gateway until the ingress registration seam"},
 	{From: "domains/ecs", To: "domains/ec2/ebs/metadata", Debt: "close-out: handlers/ecs (ADR-0004 blocked-dependency list; inventory-ecs boundary imports)", Reason: "legacy ECS calls another domain's implementation as an internal API"},
 	{From: "domains/ecs", To: "domains/ec2/eip", Debt: "close-out: handlers/ecs (ADR-0004 blocked-dependency list; inventory-ecs boundary imports)", Reason: "legacy ECS calls another domain's implementation as an internal API"},
 	{From: "domains/ecs", To: "domains/ec2/instancetypes", Debt: "close-out: handlers/ecs (ADR-0004 blocked-dependency list; inventory-ecs boundary imports)", Reason: "legacy ECS calls another domain's implementation as an internal API"},
@@ -160,6 +160,4 @@ var allowlist = []allowance{
 	{From: "domains/rds/instance", To: "runtime/compute/vm", Debt: "close-out: handlers/rds (ADR-0004 blocked-dependency list; inventory-rds boundary imports)", Reason: "legacy RDS calls another domain's or runtime implementation directly"},

 	{From: "domains/sts", To: "domains/iam", Debt: "IAM implementation consumers in STS, Ochre and ELBv2", Reason: "STS reads IAM service, trust-policy and OIDC-provider state; needs an IAM capability"},
-
-	{From: "operator/cli", To: "gateway", Debt: "operator/cli (dependency)", Reason: "command tree imports implementations; needs public capabilities"},
 }
@@ -11,7 +11,6 @@ import (
 // completed move whose line must be deleted.
 var frozenLegacyPackages = []string{
 	"gateway",
-	"gateway/spx",
 }

 // retiredLegacyRoots were legacy roots whose every package has moved to its
@@ -0,0 +1,96 @@
+#!/bin/sh
+mount -t proc  proc  /proc
+mount -t sysfs sys   /sys
+mount -t devtmpfs dev /dev
+
+# virtio_mmio is built-in (CONFIG_VIRTIO_MMIO=y) so virtio-net devices are
+# already enumerated by the time PID 1 runs — no modprobe needed.
+# qemu_fw_cfg and virtio_net are =m in stock Alpine linux-virt, so load them.
+#
+# modprobe needs a usable modules.dep; when the image is built on a host whose
+# depmod can't produce one, busybox modprobe silently no-ops and the module
+# never loads — fw_cfg then never appears and init panics. insmod by explicit
+# path needs no modules.dep, so fall back to it. Handles .ko and .ko.gz.
+load_mod() {
+    modprobe "$1" 2>/dev/null && return 0
+    for ko in $(find /lib/modules -name "$1.ko" -o -name "$1.ko.gz" 2>/dev/null); do
+        insmod "$ko" 2>/dev/null && return 0
+        case "$ko" in
+            *.gz) gunzip -c "$ko" >/tmp/mod.ko 2>/dev/null \
+                  && insmod /tmp/mod.ko 2>/dev/null && return 0 ;;
+        esac
+    done
+    return 1
+}
+load_mod qemu_fw_cfg || load_mod fw_cfg_sysfs || true
+# virtio_net depends on net_failover, which depends on failover. The load_mod
+# insmod fallback loads a single .ko with no dependency resolution, and
+# modules.dep is absent when the image is built on a kmod-less host (depmod
+# skipped), so modprobe can't resolve the chain either. Load it explicitly
+# bottom-up so the NIC comes up regardless of how the image was built —
+# without this the agent has no network and the LB hangs in provisioning.
+load_mod failover || true
+load_mod net_failover || true
+load_mod virtio_net || true
+
+NETCFG=/sys/firmware/qemu_fw_cfg/by_name/opt/spinifex/netcfg/raw
+i=0
+while [ "$i" -lt 25 ]; do
+    [ -f "$NETCFG" ] && break
+    sleep 0.04
+    i=$((i + 1))
+done
+
+if [ ! -f "$NETCFG" ]; then
+    # Dump host/guest fw_cfg state to the serial console before init exits so a
+    # "netcfg not available" panic is diagnosable from the captured console log
+    # (module-not-loaded vs qemu-not-exposing-by_name vs wrong machine type).
+    echo "[init] FATAL: fw_cfg netcfg not available" >&2
+    echo "[init] dmi: vendor=$(cat /sys/class/dmi/id/sys_vendor 2>/dev/null) product=$(cat /sys/class/dmi/id/product_name 2>/dev/null) bios=$(cat /sys/class/dmi/id/bios_version 2>/dev/null)" >&2
+    echo "[init] qemu_fw_cfg dir present: $([ -d /sys/firmware/qemu_fw_cfg ] && echo yes || echo NO)" >&2
+    echo "[init] modules: $(lsmod 2>/dev/null | awk 'NR>1{print $1}' | tr '\n' ',')" >&2
+    echo "[init] by_name listing:" >&2
+    ls -la /sys/firmware/qemu_fw_cfg/by_name/ 2>&1 | sed 's/^/[init]   /' >&2
+    ls -la /sys/firmware/qemu_fw_cfg/by_name/opt/ 2>&1 | sed 's/^/[init]   /' >&2
+    echo "[init] dmesg fw_cfg:" >&2
+    dmesg 2>/dev/null | grep -i fw_cfg | sed 's/^/[init]   /' >&2
+    exit 1
+fi
+
+. "$NETCFG"
+
+for n in 0 1 2 3 4 5; do
+    eval mac=\$NIC${n}_MAC
+    [ -z "$mac" ] && continue
+
+    iface=""
+    for attempt in 1 2 3 4 5; do
+        iface=$(ip -o link | awk -v m="$mac" 'tolower($0) ~ tolower(m) {print $2}' | tr -d :)
+        [ -n "$iface" ] && break
+        sleep 0.02
+    done
+
+    if [ -z "$iface" ]; then
+        echo "[init] WARNING: no interface found for MAC $mac (NIC$n), skipping" >&2
+        continue
+    fi
+
+    eval cidr=\$NIC${n}_CIDR
+    eval gw=\$NIC${n}_GW
+    eval default=\$NIC${n}_DEFAULT
+    eval rdst=\$NIC${n}_ROUTE_DST
+    eval rvia=\$NIC${n}_ROUTE_VIA
+
+    ip link set "$iface" up
+    [ -n "$cidr" ] && ip addr add "$cidr" dev "$iface"
+    [ "$default" = "1" ] && [ -n "$gw" ] && ip route add default via "$gw" dev "$iface"
+    [ -n "$rdst" ] && [ -n "$rvia" ] && ip route add "$rdst" via "$rvia" dev "$iface"
+done
+
+# The store image takes no lb-agent environment. A CA bundle is optional here:
+# nothing in this image dials out yet.
+CA_CERT=/sys/firmware/qemu_fw_cfg/by_name/opt/spinifex/ca-cert/raw
+mkdir -p /etc/ssl/certs
+[ -f "$CA_CERT" ] && cp "$CA_CERT" /etc/ssl/certs/ca-certificates.crt
+
+exec /sbin/init
@@ -0,0 +1,10 @@
+# The power button loads here, not in /init: busybox init has its SIGUSR2
+# (poweroff) handler installed before sysinit runs, so a press can never reach
+# the pre-exec shell PID 1. A press before this line is lost; the host re-sends.
+::sysinit:/sbin/modprobe tiny_power_button power_signal=12
+::sysinit:/usr/local/lib/sqs-store/sqs-store-prepare
+::respawn:/usr/local/lib/sqs-store/sqs-store-run
+::ctrlaltdel:/sbin/reboot
+# One action: the store stops and its disk unmounts first, and the generic
+# umount -a -r runs only if the store is gone.
+::shutdown:/usr/local/lib/sqs-store/sqs-store-stop
@@ -0,0 +1,215 @@
+# sqs-store-lib.sh — shared by the SQS store guest's prepare, run and stop
+# steps. Sourced, never executed. Every path and command is overridable so the
+# host test can drive the same code against a fake fw_cfg, sysfs and tools.
+#
+# Guest contract, read from fw_cfg under opt/spinifex/sqs-store/:
+#   data-serial  virtio-blk serial of the already-formatted ext4 data disk
+#   listen-addr  IPv4 address on a configured guest NIC; the only listener
+#   client-nkey  NATS user NKey public key allowed to connect; never a seed
+# Anything absent, malformed or ambiguous refuses the store. Nothing formats.
+
+SQS_FWCFG=${SQS_FWCFG:-/sys/firmware/qemu_fw_cfg/by_name/opt/spinifex/sqs-store}
+SQS_SYSBLOCK=${SQS_SYSBLOCK:-/sys/block}
+SQS_DEVDIR=${SQS_DEVDIR:-/dev}
+# The host test cannot create block nodes, so it relaxes this to -e.
+SQS_NODE_TEST=${SQS_NODE_TEST:--b}
+SQS_RUN=${SQS_RUN:-/run/sqs-store}
+SQS_DATA=${SQS_DATA:-/data}
+SQS_CONSOLE=${SQS_CONSOLE:-/dev/console}
+SQS_MOUNTS=${SQS_MOUNTS:-/proc/mounts}
+SQS_PORT=${SQS_PORT:-4222}
+SQS_STOP_BUDGET_TENTHS=${SQS_STOP_BUDGET_TENTHS:-120}
+NATS_BIN=${NATS_BIN:-/usr/local/bin/nats-server}
+E2FSCK=${E2FSCK:-e2fsck}
+MOUNT=${MOUNT:-mount}
+UMOUNT=${UMOUNT:-umount}
+IP=${IP:-ip}
+
+sqs_log() { echo "[sqs-store] $*" >> "$SQS_CONSOLE"; }
+sqs_fatal() { sqs_log "FATAL: $*"; return 1; }
+
+# sqs_fwcfg <name>: print one fw_cfg value, refusing a missing, empty, oversized
+# or multi-token file. The value is not echoed on failure.
+sqs_fwcfg() {
+    f="$SQS_FWCFG/$1/raw"
+    [ -f "$f" ] || { sqs_fatal "fw_cfg $1 not supplied"; return 1; }
+    [ "$(wc -c < "$f")" -le 128 ] || { sqs_fatal "fw_cfg $1 is oversized"; return 1; }
+    v=$(cat "$f")
+    case "$v" in
+        "" | *[[:space:]]*) sqs_fatal "fw_cfg $1 must be exactly one value"; return 1 ;;
+    esac
+    printf '%s' "$v"
+}
+
+# A volume serial is the volume ID with dashes stripped: vol + 17 hex.
+sqs_valid_serial() {
+    [ "${#1}" -eq 20 ] || return 1
+    case "$1" in vol*) ;; *) return 1 ;; esac
+    case "${1#vol}" in *[!0-9a-f]*) return 1 ;; esac
+}
+
+# A NATS user public NKey: "U" then 55 base32 characters. The checksum is
+# left to the pinned server's own config test.
+sqs_valid_user_nkey() {
+    [ "${#1}" -eq 56 ] || return 1
+    case "$1" in U*) ;; *) return 1 ;; esac
+    case "$1" in *[!A-Z2-7]*) return 1 ;; esac
+}
+
+sqs_valid_ipv4() {
+    case "$1" in *[!0-9.]* | .* | *. | *..*) return 1 ;; esac
+    old_ifs=$IFS; IFS=.
+    # shellcheck disable=SC2086 # split on dots on purpose
+    set -- $1
+    IFS=$old_ifs
+    [ "$#" -eq 4 ] || return 1
+    for o in "$@"; do
+        [ "${#o}" -le 3 ] && [ "$o" -le 255 ] || return 1
+        case "$o" in 0?*) return 1 ;; esac
+    done
+}
+
+# sqs_find_disk <serial>: print the /dev node of the one block device whose
+# sysfs serial matches exactly. Zero or several matches refuse.
+sqs_find_disk() {
+    found="" n=0
+    for b in "$SQS_SYSBLOCK"/*; do
+        [ -r "$b/serial" ] || continue
+        [ "$(cat "$b/serial")" = "$1" ] || continue
+        found=${b##*/}
+        n=$((n + 1))
+    done
+    [ "$n" -eq 1 ] || { sqs_fatal "expected one disk with serial $1, found $n"; return 1; }
+    test "$SQS_NODE_TEST" "$SQS_DEVDIR/$found" || { sqs_fatal "disk $found has no block device node"; return 1; }
+    printf '%s' "$SQS_DEVDIR/$found"
+}
+
+# sqs_has_addr <ipv4>: the address is assigned to a non-loopback interface.
+sqs_has_addr() {
+    "$IP" -o -4 addr show 2>/dev/null \
+        | awk -v a="$1" '$2 != "lo" { split($4, p, "/"); if (p[1] == a) f = 1 } END { exit !f }'
+}
+
+# sqs_alive <pid>: running or stopped, not exited (a zombie counts as exited).
+sqs_alive() { [ -r "/proc/$1/status" ] && ! grep -q '^State:[[:space:]]*Z' "/proc/$1/status" 2>/dev/null; }
+
+# sqs_starttime <pid>: kernel start time (stat field 22). It survives exec, and
+# pairing it with the pid reduces PID-reuse risk; without a pidfd the check and
+# the signal stay separate steps, so it is not an absolute guarantee.
+sqs_starttime() { sed 's/.*) //' "/proc/$1/stat" 2>/dev/null | cut -d' ' -f20; }
+
+# sqs_same <pid> <starttime>: the recorded process is still alive.
+sqs_same() { [ -n "$2" ] && sqs_alive "$1" && [ "$(sqs_starttime "$1")" = "$2" ]; }
+
+# sqs_prepare: validate the contract, check and mount the data disk, write the
+# server config and mark the store ready. Returns non-zero, unready, on any refusal.
+sqs_prepare() {
+    rm -f "$SQS_RUN/ready"
+    mkdir -p "$SQS_RUN"
+    serial=$(sqs_fwcfg data-serial) || return 1
+    sqs_valid_serial "$serial" || { sqs_fatal "data-serial is not a volume serial"; return 1; }
+    addr=$(sqs_fwcfg listen-addr) || return 1
+    sqs_valid_ipv4 "$addr" || { sqs_fatal "listen-addr is not an IPv4 address"; return 1; }
+    nkey=$(sqs_fwcfg client-nkey) || return 1
+    sqs_valid_user_nkey "$nkey" || { sqs_fatal "client-nkey is not a user public NKey"; return 1; }
+    sqs_has_addr "$addr" || { sqs_fatal "listen-addr $addr is not on any guest NIC"; return 1; }
+
+    # The server checks the config, including the NKey checksum, before the
+    # disk is touched.
+    umask 077
+    cat > "$SQS_RUN/nats.conf" <<EOF
+listen: "$addr:$SQS_PORT"
+server_name: "sqs-store"
+jetstream {
+  store_dir: "$SQS_DATA"
+  sync_interval: always
+}
+authorization {
+  users = [ { nkey: "$nkey" } ]
+}
+EOF
+    if ! "$NATS_BIN" -t -c "$SQS_RUN/nats.conf" > "$SQS_RUN/config-test.out" 2>&1; then
+        sqs_fatal "server config test failed: $(sed 's/U[A-Z2-7]\{55\}/<nkey>/g' "$SQS_RUN/config-test.out" | tr '\n' ' ')"
+        return 1
+    fi
+
+    dev=$(sqs_find_disk "$serial") || return 1
+    rc=0
+    "$E2FSCK" -p "$dev" > "$SQS_RUN/fsck.out" 2>&1 || rc=$?
+    sqs_log "e2fsck -p $dev rc=$rc"
+    # 0 clean, 1 corrected: proceed. Anything higher, including a disk that is
+    # not ext4 at all, refuses; the guest never formats.
+    [ "$rc" -le 1 ] || { sqs_fatal "e2fsck refused $dev: $(tr '\n' ' ' < "$SQS_RUN/fsck.out")"; return 1; }
+    mkdir -p "$SQS_DATA"
+    "$MOUNT" -t ext4 -o nodev,nosuid,noexec "$dev" "$SQS_DATA" || { sqs_fatal "mount $dev failed"; return 1; }
+    : > "$SQS_RUN/ready"
+    sqs_log "prepared: $dev mounted at $SQS_DATA, config written for $addr:$SQS_PORT; server not started yet"
+}
+
+# sqs_run: the respawned supervisor body. Its identity is recorded before the
+# stopping check: a stop that misses it set stopping first, so this parks; a
+# stop that reads it signals this process whether or not it has exec'd yet.
+sqs_run() {
+    mkdir -p "$SQS_RUN"
+    echo "$$ $(sqs_starttime $$)" > "$SQS_RUN/pid"
+    if [ -e "$SQS_RUN/stopping" ]; then
+        sqs_log "run: shutdown in progress, not starting"
+        return 2
+    fi
+    if [ ! -e "$SQS_RUN/ready" ]; then
+        sqs_log "run: store not prepared, not starting"
+        return 2
+    fi
+    sqs_log "run: starting pid=$$"
+    exec "$NATS_BIN" -c "$SQS_RUN/nats.conf"
+}
+
+# sqs_stop: first shutdown action. Block any respawn, SIGTERM the recorded run
+# process (a server, or a start not yet exec'd), SIGKILL only after the budget,
+# then sync and unmount. Each signal rechecks the identity; a survivor keeps the disk mounted.
+sqs_stop() {
+    mkdir -p "$SQS_RUN"
+    : > "$SQS_RUN/stopping"
+    sqs_log "stop: begin"
+    rec=$(cat "$SQS_RUN/pid" 2>/dev/null || true)
+    pid=${rec%% *} st=${rec#* }
+    [ "$st" != "$rec" ] || st=""
+    if [ -n "$pid" ] && sqs_same "$pid" "$st"; then
+        kill -TERM "$pid" 2>/dev/null || true
+        sqs_log "stop: SIGTERM sent to run pid=$pid"
+        i=0
+        while sqs_same "$pid" "$st" && [ "$i" -lt "$SQS_STOP_BUDGET_TENTHS" ]; do sleep 0.1; i=$((i + 1)); done
+        if sqs_same "$pid" "$st"; then
+            sqs_log "stop: run pid=$pid still running after $((SQS_STOP_BUDGET_TENTHS / 10))s, SIGKILL"
+            kill -KILL "$pid" 2>/dev/null || true
+            j=0
+            while sqs_same "$pid" "$st" && [ "$j" -lt 30 ]; do sleep 0.1; j=$((j + 1)); done
+        fi
+        if sqs_same "$pid" "$st"; then
+            sqs_log "stop: run pid=$pid STILL ALIVE, leaving $SQS_DATA mounted"
+            return 1
+        fi
+        sqs_log "stop: store exited after $((i / 10)).$((i % 10))s"
+    else
+        sqs_log "stop: no running store"
+    fi
+    sync
+    if grep -q " $SQS_DATA " "$SQS_MOUNTS" 2>/dev/null; then
+        rc=0
+        "$UMOUNT" "$SQS_DATA" || rc=$?
+        sqs_log "stop: umount $SQS_DATA rc=$rc"
+    fi
+}
+
+# sqs_shutdown: the only shutdown action. The generic read-only remount runs
+# only after the store stopped; a surviving writer keeps every mount as it is.
+sqs_shutdown() {
+    if ! sqs_stop; then
+        sync
+        sqs_log "shutdown: store survived, skipping umount -a -r"
+        return 1
+    fi
+    rc=0
+    "$UMOUNT" -a -r || rc=$?
+    echo "[init] shutdown: umount -a -r rc=$rc" >> "$SQS_CONSOLE"
+}
@@ -0,0 +1,9 @@
+#!/bin/sh
+# sysinit: load the disk drivers, then validate the guest contract, check and
+# mount the data disk and write the server config. A refusal leaves the store
+# unprepared, so the run step never starts it.
+. /usr/local/lib/sqs-store/sqs-store-lib.sh
+modprobe virtio_blk 2>/dev/null || true
+modprobe ext4 2>/dev/null || true
+sqs_prepare || sqs_log "store refused; it will not start"
+exit 0
@@ -0,0 +1,6 @@
+#!/bin/sh
+# Respawned by init. A store that must not run parks here instead of exiting,
+# so init does not respawn it in a tight loop.
+. /usr/local/lib/sqs-store/sqs-store-lib.sh
+sqs_run
+exec sleep 2147483647
@@ -0,0 +1,5 @@
+#!/bin/sh
+# The only shutdown action: stop the store and unmount its disk, then run the
+# generic read-only remount only if the store is gone.
+. /usr/local/lib/sqs-store/sqs-store-lib.sh
+sqs_shutdown
@@ -1,57 +0,0 @@
-# spinifex-unit-version: 1
-[Unit]
-Description=Spinifex SQS Message Store (dedicated JetStream)
-PartOf=spinifex.target
-ConditionPathExists=/etc/spinifex/sqs-store/sqs-store.conf
-
-[Service]
-EnvironmentFile=-/etc/spinifex/telemetry.env
-Type=simple
-Slice=spinifex-system.slice
-User=spinifex-sqsstore
-Group=spinifex
-ExecStart=/usr/local/bin/spx service sqs-store start
-Restart=on-failure
-RestartSec=5
-# JetStream store shutdown and raft stepdown have no internal deadline.
-TimeoutStopSec=20
-OOMScoreAdjust=-500
-# Soft Go heap limit for the store; MemoryHigh/Max bound a runaway on top.
-Environment=GOMEMLIMIT=512MiB
-MemoryHigh=10%
-MemoryMax=20%
-Environment=SPINIFEX_CONFIG_PATH=/etc/spinifex/sqs-store/sqs-store.conf
-
-# Filesystem access
-NoNewPrivileges=yes
-ProtectSystem=strict
-PrivateDevices=yes
-ReadOnlyPaths=/etc/spinifex/sqs-store /etc/spinifex/server.pem /etc/spinifex/server.key /etc/spinifex/ca.pem
-ReadWritePaths=/var/lib/spinifex/sqs-store /var/log/spinifex
-
-# Hardening
-ProtectHome=yes
-PrivateTmp=yes
-RestrictSUIDSGID=yes
-LockPersonality=yes
-RestrictRealtime=yes
-RestrictNamespaces=yes
-ProtectClock=yes
-ProtectKernelTunables=yes
-ProtectKernelModules=yes
-ProtectKernelLogs=yes
-ProtectControlGroups=yes
-ProtectProc=invisible
-SystemCallArchitectures=native
-MemoryDenyWriteExecute=yes
-RestrictAddressFamilies=AF_INET AF_INET6 AF_UNIX AF_NETLINK
-CapabilityBoundingSet=
-ProtectHostname=yes
-KeyringMode=private
-RemoveIPC=yes
-LimitCORE=0
-LimitNOFILE=65536
-LimitNPROC=4096
-
-[Install]
-WantedBy=spinifex.target
@@ -1,9 +1,9 @@
-# spinifex-unit-version: 3
+# spinifex-unit-version: 4
 [Unit]
 Description=Spinifex Infrastructure Platform
 After=network-online.target
 Wants=network-online.target
-Wants=spinifex-nats.service spinifex-sqs-store.service spinifex-predastore.service spinifex-northstar.service spinifex-viperblock.service spinifex-daemon.service spinifex-awsgw.service spinifex-vpcd.service spinifex-ui.service spinifex-qmp-collector.service spinifex-shutdown.service spinifex-nats-watchdog.timer ovsdb-server.service ovs-vswitchd.service ovn-controller.service
+Wants=spinifex-nats.service spinifex-predastore.service spinifex-northstar.service spinifex-viperblock.service spinifex-daemon.service spinifex-awsgw.service spinifex-vpcd.service spinifex-ui.service spinifex-qmp-collector.service spinifex-shutdown.service spinifex-nats-watchdog.timer ovsdb-server.service ovs-vswitchd.service ovn-controller.service

 [Install]
 WantedBy=multi-user.target
@@ -10,6 +10,8 @@ import (
 	"github.com/mulgadc/spinifex/internal/awsmodel"
 	acmawsapi "github.com/mulgadc/spinifex/spinifex/domains/acm/awsapi"
 	awsapi "github.com/mulgadc/spinifex/spinifex/domains/ecr/awsapi"
+	elbv2awsapi "github.com/mulgadc/spinifex/spinifex/domains/elbv2/awsapi"
+	rdsawsapi "github.com/mulgadc/spinifex/spinifex/domains/rds/awsapi"
 	route53awsapi "github.com/mulgadc/spinifex/spinifex/domains/route53/awsapi"
 	"github.com/mulgadc/spinifex/spinifex/gateway"
 	"github.com/mulgadc/spinifex/spinifex/ingress/aws/dispatch"
@@ -54,6 +56,8 @@ func compareAll() ([]awsmodel.OperationCoverage, error) {
 		awsapi.ServiceName:        awsapi.OperationInventory(),
 		acmawsapi.ServiceName:     acmawsapi.OperationInventory(),
 		route53awsapi.ServiceName: route53awsapi.OperationInventory(),
+		elbv2awsapi.InventoryName: elbv2awsapi.OperationInventory(),
+		rdsawsapi.ServiceName:     rdsawsapi.OperationInventory(),
 	})
 	// The loop below walks the embedded models, so a served service with no
 	// model would otherwise drop out of the report without a word.
@@ -8,3 +8,8 @@ administration. It is separate from AWS tenant-facing service contracts.
 roles and Reed-Solomon configuration. It is an operational status contract,
 not a generic provider abstraction; do not change its subject or JSON shape
 incompatibly. Add a new version when that is required.
+
+`admin.go` holds the `/admin/<Method>` request and response bodies the private admin surface shares with the CLI.
+`AdminMethodNames` is the list of callable methods, and the gateway router builds its accept set from it.
+`AccountQuotaRequest`, `AccountQuotaResponse` and `DeleteAccountResponse` stay in the gateway until the `/admin` handlers move, because they embed types the contracts layer may not import.
+Do not change a field name, JSON tag or omitempty incompatibly; add a new version when that is required.
@@ -0,0 +1,76 @@
+package operatorv1
+
+import "slices"
+
+// adminMethodNames is the callable /admin/<Method> list. The gateway router
+// builds its accept set from it, so a name cannot be served without being listed.
+var adminMethodNames = []string{
+	"CreateAccount",
+	"DeleteAccount",
+	"DescribeAccountDeletion",
+	"GetAccountQuota",
+	"ListAccounts",
+	"PutAccountQuota",
+}
+
+// AdminMethodNames returns the callable /admin/<Method> names in sorted order.
+// Tooling that mints a credential grants these by name, so the list has to be
+// the one the router serves rather than a copy that can drift out of step.
+func AdminMethodNames() []string {
+	return slices.Sorted(slices.Values(adminMethodNames))
+}
+
+// Deletion job states.
+const (
+	DeletionStateRunning   = "RUNNING"
+	DeletionStateCompleted = "COMPLETED"
+	DeletionStateFailed    = "FAILED"
+)
+
+// CreateAccountRequest is the /admin/CreateAccount request body.
+type CreateAccountRequest struct {
+	Name        string `json:"name"`
+	ClientToken string `json:"clientToken"`
+	Source      string `json:"source,omitempty"`
+}
+
+// CreateAccountResponse is the /admin/CreateAccount success body. It is stored
+// verbatim (with the secret encrypted) so a replay returns an identical answer.
+type CreateAccountResponse struct {
+	AccountID       string `json:"accountId"`
+	AccountName     string `json:"accountName"`
+	AdminUser       string `json:"adminUser"`
+	AccessKeyID     string `json:"accessKeyId"`
+	SecretAccessKey string `json:"secretAccessKey"`
+	DefaultVpcID    string `json:"defaultVpcId"`
+}
+
+// DeleteAccountRequest is the /admin/DeleteAccount request body.
+type DeleteAccountRequest struct {
+	AccountID   string `json:"accountId"`
+	AccountName string `json:"accountName"`
+	ClientToken string `json:"clientToken"`
+	Force       bool   `json:"force,omitempty"`
+	DryRun      bool   `json:"dryRun,omitempty"`
+}
+
+// DescribeAccountDeletionRequest is the /admin/DescribeAccountDeletion body.
+type DescribeAccountDeletionRequest struct {
+	AccountID string `json:"accountId"`
+}
+
+// AccountSummary is one row of /admin/ListAccounts. It carries no credential
+// material: the caller can already create and delete accounts, but that is no
+// reason for a listing to hand out keys.
+type AccountSummary struct {
+	AccountID   string `json:"accountId"`
+	AccountName string `json:"accountName"`
+	Status      string `json:"status"`
+	CreatedAt   string `json:"createdAt,omitempty"`
+}
+
+// ListAccountsResponse is the /admin/ListAccounts success body.
+type ListAccountsResponse struct {
+	Accounts []AccountSummary `json:"accounts"`
+	Count    int              `json:"count"`
+}
@@ -0,0 +1,65 @@
+package operatorv1
+
+import (
+	"encoding/json"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+func goldenJSON(t *testing.T, v any, want string) {
+	t.Helper()
+	data, err := json.Marshal(v)
+	require.NoError(t, err)
+	require.JSONEq(t, want, string(data))
+}
+
+func TestAdminWire_CreateAccountRequest(t *testing.T) {
+	goldenJSON(t, CreateAccountRequest{Name: "acme", ClientToken: "tok-1", Source: "cli"},
+		`{"name":"acme","clientToken":"tok-1","source":"cli"}`)
+	goldenJSON(t, CreateAccountRequest{Name: "acme", ClientToken: "tok-1"},
+		`{"name":"acme","clientToken":"tok-1"}`)
+}
+
+func TestAdminWire_CreateAccountResponse(t *testing.T) {
+	goldenJSON(t, CreateAccountResponse{
+		AccountID: "000000000042", AccountName: "acme", AdminUser: "admin",
+		AccessKeyID: "AKIAEXAMPLE", SecretAccessKey: "secret", DefaultVpcID: "vpc-1",
+	}, `{"accountId":"000000000042","accountName":"acme","adminUser":"admin","accessKeyId":"AKIAEXAMPLE","secretAccessKey":"secret","defaultVpcId":"vpc-1"}`)
+}
+
+func TestAdminWire_DeleteAccountRequest(t *testing.T) {
+	goldenJSON(t, DeleteAccountRequest{AccountID: "000000000042", AccountName: "acme", ClientToken: "tok-2", Force: true, DryRun: true},
+		`{"accountId":"000000000042","accountName":"acme","clientToken":"tok-2","force":true,"dryRun":true}`)
+	goldenJSON(t, DeleteAccountRequest{AccountID: "000000000042", AccountName: "acme", ClientToken: "tok-2"},
+		`{"accountId":"000000000042","accountName":"acme","clientToken":"tok-2"}`)
+}
+
+func TestAdminWire_DescribeAccountDeletionRequest(t *testing.T) {
+	goldenJSON(t, DescribeAccountDeletionRequest{AccountID: "000000000042"}, `{"accountId":"000000000042"}`)
+}
+
+func TestAdminWire_DeletionStates(t *testing.T) {
+	require.Equal(t, "RUNNING", DeletionStateRunning)
+	require.Equal(t, "COMPLETED", DeletionStateCompleted)
+	require.Equal(t, "FAILED", DeletionStateFailed)
+}
+
+func TestAdminWire_ListAccountsResponse(t *testing.T) {
+	goldenJSON(t, ListAccountsResponse{
+		Accounts: []AccountSummary{
+			{AccountID: "000000000042", AccountName: "acme", Status: "ACTIVE", CreatedAt: "2026-01-02T03:04:05Z"},
+			{AccountID: "000000000043", AccountName: "bolt", Status: "TERMINATING"},
+		},
+		Count: 2,
+	}, `{"accounts":[
+		{"accountId":"000000000042","accountName":"acme","status":"ACTIVE","createdAt":"2026-01-02T03:04:05Z"},
+		{"accountId":"000000000043","accountName":"bolt","status":"TERMINATING"}],"count":2}`)
+}
+
+func TestAdminWire_AdminMethodNames(t *testing.T) {
+	require.Equal(t, []string{
+		"CreateAccount", "DeleteAccount", "DescribeAccountDeletion",
+		"GetAccountQuota", "ListAccounts", "PutAccountQuota",
+	}, AdminMethodNames())
+}
@@ -243,6 +243,22 @@ Hot-attaching a volume to a running direct-boot microVM is not supported, becaus

 This covers only requests that are already mounted when the instance starts; nothing yet supplies data volumes to a system instance, and the path is verified with fake-QEMU unit tests, not with a real EBS volume in a guest.

+#### Volume leases
+
+Before opening an engine on a volume, the Viperblock worker takes a lease: one entry per volume in the `VIPERBLOCK_VOLUME_LEASES` JetStream KV bucket, which expires 45 seconds after its last write (`providers/ebs/viperblock/volume_lease.go`).
+A renewal becomes eligible once the last confirmed write is 10 seconds old, and the holder checks for that periodically, so the rewrite lands some time after that point.
+Each rewrite is conditioned on the revision the holder last wrote, and a refused write that finds the entry gone or naming another owner or generation fences the export as a takeover.
+It may keep writing for 40 seconds after the last write the store acknowledged, measured from when that write was sent, because the server starts the entry's expiry when it applies the write and never earlier.
+A slow acknowledgement therefore shortens the remaining validity rather than extending it, and an acquisition acknowledged after its validity has already run out is refused as retryable.
+
+For a holder that cannot renew, a timer is armed for the validity deadline, leaving a nominal 5 seconds before the entry can expire; when it fires, the export's nbdkit is fenced with SIGKILL (`volume_fence.go`).
+The fence runs on its own goroutine, so Go scheduling, that goroutine and process termination all delay the kill, and enough delay can use up the whole margin; a kill held in uninterruptible sleep, which the fence waits up to 30 seconds for, is the extreme case.
+Once a lease is lost or has lapsed it stays lost: a renewal that succeeds afterwards is ignored, and another open on the same node cannot share it while the fence runs.
+
+These rules bound only the holder's own clock.
+The storage backend does not reject a stale writer, so a paused or partitioned host that resumes after its validity can still write; generation enforcement at the write boundary is required by Q-24 in the umbrella repository's `docs/specs/viperblock/rfc3-1_overview.md` and is not implemented.
+An export that survives a daemon restart writes without a renewed lease until recovery adopts it, and a survivor that recovery cannot adopt while a guest still uses it is left running without a lease or a fence.
+
 ### Predastore (S3)

 Object storage used for:
@@ -274,7 +290,7 @@ storage-provider contract.

 `domains/sqs/message` defines the store-neutral interface the SQS domain reaches its message store through (ADR-0011:S6), with a JetStream adapter in `domains/sqs/message/jetstream` and a shared conformance suite in `domains/sqs/message/conformance`.
 It is an internal primitive: nothing outside the SQS domain uses it, no daemon or service is wired to it, and it does not itself deliver an SQS API.
-The dedicated JetStream server ADR-0011 describes is not deployed by this code.
+The dedicated JetStream server ADR-0011 describes is not deployed by this code, and Spinifex ships no host service, unit, user or firewall rule for it; the guest that could run it is described under *SQS store system instance (not wired)*.

 The interface covers durable and idempotent append, body and metadata reads, conditional state writes and deletes, per-namespace counts, namespace enumeration, an ordered change feed, a bounded paginated scan, namespace deletion and a sweep of tombstoned bodies.
 A namespace carries an immutable message-shard and state-shard allocation chosen when the queue is created, so a later change in shard counts never moves an existing queue.
@@ -293,6 +309,40 @@ If that consumer delivers nothing for about two seconds while the stream holds a

 Coverage is unit and conformance tests on in-process R1 and R3 servers; there is no end-to-end or multi-host validation of this code, and earlier multi-host evidence predates the final feed.

+### SQS store system instance (not wired)
+
+Spinifex can build, install, launch, restart and give storage to a dedicated SQS store microVM, but nothing calls any of it.
+No SQS API, domain service, admin command or daemon start-up path creates a store, so none of this is in production use.
+
+The guest image is built by `make build-sqs-store-microvm-image`, separate from the load balancer image, and bundles the `nats-server` version pinned in `go.mod` plus the ext4 tools.
+Release tarballs carry it as `microvm-sqs-store/`, and `setup.sh` installs it to `/usr/share/spinifex/microvm/sqs-store/`.
+A release without a complete image removes any previously installed one, so a launch is then refused rather than booting a stale image.
+
+At boot the guest reads three firmware config values: the data disk's serial, an IPv4 address on one of its interfaces, and one NATS user NKey public key.
+It mounts exactly one matching disk at `/data` after `e2fsck -p`, refuses a missing, duplicate, unformatted or unrepairable disk, and never formats.
+JetStream keeps its data on `/data` with `sync_interval: always` (ADR-0011:S3), and the server listens only on that address and accepts only that key.
+On shutdown the store is stopped before `/data` is unmounted; a store that survives SIGKILL leaves every mount in place.
+
+A launch uses the versioned subject `system.LaunchSQSStore.v1`, which a daemon without this code does not serve, so it fails with no responder instead of booting another guest.
+The legacy `system.LaunchInstance` path refuses the store kind on both send and receive.
+Every input check runs before capacity is reserved, and the data volume must be an available system-account volume.
+It is mounted through the normal EBS path and detached, never deleted, on terminate, which is the protection standard EBS gives today and nothing stronger.
+CPU mitigations stay enabled for the store, unlike the load balancer.
+A successful launch means the VM is running, not that NATS is ready.
+
+The instance record keeps the store's non-secret guest settings so a restart on the node that launched it rewrites the firmware config files.
+A relaunch on any other node is refused, and cross-node recovery never claims a store.
+Both checks exist only in daemons with this code: an older daemon can still claim and relaunch a store elsewhere, and an explicit start served by another node claims the instance before its relaunch is refused.
+
+`domains/sqs/storevolume` prepares a fresh data volume, because the guest never formats.
+It checks for `mkfs.ext4` and `qemu-img` before allocating, creates the volume in `creating`, writes an empty ext4 image to it over NBD on the local node, unpublishes it, and only then makes it `available`.
+While a volume is `creating`, attach, delete, modify, force-detach and snapshot are refused.
+A volume is deleted only after a confirmed unpublish; any publish error, an unconfirmed unpublish or a failed release keeps it and reports its ID, with its state reported as unknown after a failed release.
+
+Evidence is local only: unit tests, real QEMU boots against disposable scratch disks, and one run of the preparer against the `qemunbdd` provider whose volume kept acknowledged writes across a reboot.
+Viperblock, volume encryption, the EC2 volume record path, multiple nodes and host reachability of the guest address are not tested.
+Credential issuance and rotation, store placement and NATS readiness checks do not exist, and there is no failover: a lost node leaves its store unavailable.
+
 ## Configuration

 Cluster configuration (`spinifex/bootstrap/config/config.go`):
@@ -351,8 +351,16 @@ while the branch is in flight; they do not describe the umbrella repository's
 | `spinifex#1269` | `spinifex/handlers/elbv2` | `spinifex/domains/elbv2` (package `elbv2`) | A mechanical whole-package relocation under the 2026-10-26 ADR-0001 amendment: the ELBv2 control plane keeps its files, tests and current dependencies, and importers alias it `elbv2domain`. No identifier, API, NATS subject, KV bucket, metric, error string or behaviour changes. Its allowlist entries are now sourced from `domains/elbv2`. Q-30's data-plane boundary stays deferred, recorded as `agents/elbv2` (Q-30 data-plane boundary) debt. |
 | `spinifex#1271` | `spinifex/handlers/ecs` | `spinifex/domains/ecs` (package `ecs`) | A mechanical whole-package relocation: ADR-0001's `handlers/ecs` row targets `domains/ecs`, the files are already organised by resource, and resource packages are introduced only where callers need a compiler boundary. The ECS control plane keeps its files, tests and current dependencies alongside the already-extracted `domains/ecs/taskdefinition`, and importers alias it `ecsdomain`. No identifier, API, NATS subject, KV bucket, metric, error string or behaviour changes. Its allowlist entries are now sourced from `domains/ecs`. `handlers` has no package left and is a retired legacy root; R5 holds `domains/ecs` to the interim rule's ban on importing the daemon. |
 | `spinifex#1273` | `spinifex/gateway/{ecs,elbv2}` | `spinifex/domains/{ecs,elbv2}/awsapi` (`awsapi`, imported as `ecsawsapi`, `elbv2awsapi`) | A mechanical relocation of the ECS and ELBv2 AWS action adapters to their ADR-0001 domain homes, as `spinifex#1265` did for IAM, STS, EKS and RDS; APIs, NATS subjects, KV buckets and behaviour are unchanged. The root `gateway` dispatch still imports each package, recorded as `gateway` (service registration) debt, and `domains/admission/quota` keeps its live ELBv2 counter import, now of `domains/elbv2/awsapi`. |
+| `spinifex#1336` | `spinifex/gateway/tagging.go` (`Tagging_Request`, `taggingActions`, `taggingActionFromTarget`) and the `GatewayConfig.TaggingResources` field | `spinifex/domains/tagging/awsapi` (`NewRegistration`, `Deps`, `Actions`, `OperationInventory`), composed by `spinifex/runtime/roles/awsgw` | Resource Groups Tagging registers through the ADR-0002 seam like ACM and Route 53. The check order is kept: action, scope, authorization, account, body, aggregator, so a denied caller still gets `AccessDenied` before any body error and an unwired aggregator answers `ServerInternal`. The `checkPolicyResources` stand-in of `*` for every action is unchanged. The legacy switch case, the `supportedServices` entry and the `jsonErrorService` case are removed, and the registration declares the JSON 1.1 envelope. No allowlist row existed for this path. The nil-NATS check moves to the gateway-wide gate, which answers `ServiceUnavailableException` before dispatch as it did for every service. `cmd/aws-model-coverage` output is byte-identical because there is no embedded Tagging model. The characterization tests that pin the pipeline behaviour were committed first and pass unchanged against the registration; only the `withTagging` test helper changed. |
+| `spinifex#1336` | `spinifex/gateway/elbv2.go` (`ELBv2_Request`, `elbv2Actions`, `elbv2Handler`, `accountLoadBalancerLimit`) | `spinifex/domains/elbv2/awsapi` (`NewRegistration`, `Deps`, `Quota`, `ActionNames`, `OperationInventory`), `spinifex/domains/admission/quota` (`Service.LoadBalancerLimit`), `spinifex/ingress/aws/query` (`ParseArgs`), composed by `spinifex/runtime/roles/awsgw` | ELBv2 registers through the ADR-0002 seam as the first AWS Query service. The check order is kept: query parse, action, account, input parse, scope, authorization, NATS, dispatch, so a denied caller still gets `AccessDenied` before the handler runs. The gateway members ELBv2 read are now supplied as `Deps`: `NATSConn` as `Deps.NATS`, `Quota` as the `Deps.Quota` interface declared in `awsapi` because `domains/admission/quota` imports `awsapi`, `Region` as `Invocation.Region`, and the policy gate as `Invocation.Authorize`. The `DescribeAccountLimits` cap resolution moves to `Service.LoadBalancerLimit` unchanged, and a nil `Deps.Quota` enforces nothing and reports the AWS default. AWSGW now wires the registry after it builds the quota service, because the registration captures it. The registration declares the XML envelope and the inventory name `elasticloadbalancingv2`, so `cmd/aws-model-coverage` keys it explicitly and its output is byte-identical. `gateway.ParseAWSQueryArgs` delegates to `query.ParseArgs`. The legacy switch case, the `supportedServices` entry, the inventory entry and the `gateway -> domains/elbv2/awsapi` allowlist row are removed. `cmd/aws-model-coverage` imports `domains/elbv2/awsapi` for the inventory, under a fourth named-exception allowlist row whose reason is now generic for all four. Finding, not fixed here: the XML for `DescribeAccountLimits` and `DescribeLoadBalancers` lists member fields in a different order from run to run because the marshal is map-ordered, so the characterization tests pin decoded content rather than bytes. The characterization tests that pin the pipeline behaviour were committed first and pass unchanged against the registration; only the `withELBv2` and `serveELBv2` test helpers and the setup of two older gateway tests changed. |
+| `spinifex#1337` | `spinifex/gateway/spinifex.go` (`Spinifex_Request`, `spinifexAdminActions`) | `spinifex/operator/admin` (`NewRegistration`, `Deps`, `ModelAccess`, `ActionNames`), composed by `spinifex/runtime/roles/awsgw` | The `spinifex` Query service registers through the ADR-0002 seam with the XML error envelope it had. The check order is kept: query parse, action, authorization with the wildcard resource, account, the super-admin account gate, then the action, so a denied caller learns nothing about which actions exist. The gateway members the actions read are now `Deps`: `Version` and `Commit` strings, `NATSConn` as `Deps.NATS`, `RootCAs`, `DiscoverActiveNodes` as `Deps.ActiveNodes`, `BedrockAccessAdmin` as the `ModelAccess` interface, and the credential and access resolvers as `Deps.Credentials` and `Deps.Access`. The query actions need only the caller account, so the actor is not widened. The cluster-unavailable answer stays in the shared request path ahead of authorization. `/admin/{method}` moved onto the same registration in `mulga-3nbla.7`, recorded below. |
+| `spinifex#1344` | `spinifex/gateway/rds.go` (`RDS_Request`, `rdsCaller`) | `spinifex/domains/rds/awsapi` (`NewRegistration`, `Deps`, `Quota`, `CallerFromActor`, `OperationInventory`), composed by `spinifex/runtime/roles/awsgw` | RDS registers through the ADR-0002 seam with the XML envelope it had. The check order is kept: query parse, action, account, the principal-class gate for the internal agent actions, scope, authorization, NATS, quota and dispatch, so a denied caller still gets `AccessDenied` before the handler runs and no customer grant is evaluated against an agent action. The gateway members RDS read are now supplied as `Deps`: `NATSConn` as `Deps.NATS`, `ExpectedNodes` as `Deps.ExpectedNodes`, `Quota.EnforceRDSInstances` as the `Deps.Quota` interface, `Region` as `Invocation.Region` and the policy gate as `Invocation.Authorize`. A nil `Deps.Quota` skips the DB-instance cap. The caller comes from `dispatch.Actor`: the session name for an assumed-role session and the user name otherwise, and the role name from `Actor.RoleName()`, which parses the underlying role ARN and never the session name. The account comes from `Invocation.AccountID`, so a caller with no composable ARN still reaches the policy check and is refused there as before. The legacy switch case, the `supportedServices` entry, the inventory entry and the `gateway -> domains/rds/awsapi` allowlist row are removed, and `cmd/aws-model-coverage` reads the inventory under a fifth named-exception allowlist row. The characterization tests that pin the pipeline behaviour were committed first with spinifex#1336 and pass unchanged against the registration; only the `withRDS` and `serveRDS` test helpers and the setup of the direct-call RDS tests changed. |
 | `#1277` | `spinifex/accountteardown` | `spinifex/operator/accountlifecycle` (package `accountlifecycle`, imported as `accountteardown`) | A mechanical whole-package relocation under the 2026-10-27 ADR-0001 amendment: the account deletion coordinator keeps its APIs, NATS subjects, KV access and behaviour; importers keep the `accountteardown` name as an alias, so only their import paths change. Replacing its direct domain and adapter cleanup calls with owner capabilities stays deferred debt. |
 | `#1285` | `spinifex/gateway/requestaudit.go` and `slogRequestLogger` from `spinifex/gateway/gateway.go` | `spinifex/ingress/aws/audit` (package `audit`) | The first generic-ingress extraction from the root gateway: the request-audit middleware and access-log middleware move unchanged in behaviour. The gateway's authentication flow and dispatch still populate the record, now through the exported `audit.From` and `Record` setters at the same call sites; the package imports only `foundation/telemetry` and `ingress/http`. Its unit tests moved with it; the tests that run it behind SigV4 stay in `gateway`. |
+| `spinifex#1334` | `spinifex/gateway/spx` (`spx.go`, `storage.go`, `promote_image.go` and their tests) | `spinifex/operator/admin` (package `admin`) | The version, node, VM and storage-status gather helpers and the image-promote NATS request move unchanged with no shim; the `spinifex` service dispatch that calls them stays in `gateway/spinifex.go` until the `spinifex` registration slice. `operator/admin` already imports `domains/ec2/image` and `contracts/operator/v1`, so the move adds no edge and the layering ratchet needed no allowlist row. The `gateway/spx` frozen-legacy entry is removed. |
+| `spinifex#1334` | `spinifex/gateway` (`CreateAccountRequest`/`Response`, `DeleteAccountRequest`, `DescribeAccountDeletionRequest`, `DeletionState*`, `AccountSummary`, `ListAccountsResponse`, `AdminMethodNames`) | `contracts/operator/v1` (`admin.go`) | The `/admin` bodies the CLI shares with the gateway move unchanged in field names, JSON tags and omitempty, pinned by golden JSON written against the gateway types first and moved with them. `AdminMethodNames` is now owned by the contract and the gateway builds its `adminMethods` accept set from it, with a test that the router accepts exactly that list. `AccountQuotaRequest`/`AccountQuotaResponse` (embed `quota.Overrides` from `domains/admission/quota`) and `DeleteAccountResponse` (embeds `accountlifecycle.Result`) stay in `gateway`: `contracts` may import only `foundation` and `contracts`, so they are settled when the `/admin` handlers move to `operator/admin`. `accountlifecycle.Stage` serialises as an integer, not its name; the golden test pins that. |
+| `spinifex#1345` | `spinifex/gateway/admin*.go` (the `/admin/{method}` handlers, gates and body cap) and the `AccountQuotaRequest`/`AccountQuotaResponse`/`DeleteAccountResponse` wire types | `spinifex/operator/admin` (`rest.go`, `account_*.go`), served by the same `spinifex` registration | `/admin/{method}` stays a chi route in `gateway`, ahead of the service selector, but `Admin_Request` now only mints the request ID and header and hands the request to the registered entry through `dispatchRegistered`, the path every Query request takes. `dispatch.Invocation` gains `Service`, the credential-scope service, so the scope gate is the first identity gate in the dispatcher and a key signed for another service is denied exactly as before. The order is unchanged: verb and method, then the five fail-closed gates (scope, super-admin account, IAM user, identity, `spinifex:<Method>` policy), then the cluster check, then the 8 KiB body cap. The dispatcher returns errors and the route renders them as the JSON envelope with `requestId`, then logs once from the real result, so a denial or a 5xx is never logged as a completed request; the registration's `ResolveAction` labels the method, which removes the `auth.go` special case. The caller is read from `dispatch.Actor` (`UserName` for the gate and `updatedBy`). `BucketStore` and `SignupMaxAccounts` leave `GatewayConfig` for `operator/admin.Deps`, composed in `awsgw` beside `IAM`, `Quota`, `ExpectedNodes` and `NodeIDs`. The three wire types that could not enter `contracts` move to `operator/admin` with their handlers, `operator/cli` imports them from there, and the last `operator/cli` → `gateway` allowlist entry is deleted. |
+| `spinifex#1334` | `spinifex/gateway/oidc_discovery.go` (`eksAccountKV`, the discovery-document and JWKS read and construction) | `spinifex/domains/eks/cluster` (`OIDCPublisher`, `OIDCDiscoveryDocument`), composed by `spinifex/runtime/roles/awsgw` | The EKS domain now reads its own account bucket and builds the per-cluster discovery document and JWKS responses (ADR-0005:S4), and the gateway only maps the unauthenticated `/oidc/eks/...` routes onto an `OIDCPublisher` interface it is handed. Responses are byte-identical, pinned by `oidc_discovery_current_behaviour_test.go`, whose assertions are unchanged. The gateway no longer imports `domains/eks/cluster` in production code. `STS` keeps its own `ClusterJWKSResolver` (`OIDCJWKSReader`), untouched. Recorded debt, not repaired here: a stored JWKS is served verbatim with 200 even when it is not valid JSON. |
 The EC2 contract row is intentionally different from the directory moves: it creates
 a compatibility boundary. `ec2.cmd.*` retains its deployed one-token NATS
 subscription and permission shape, while callers use
@@ -438,7 +446,7 @@ These residuals were recorded by individual slices and must be resolved once the
 - `runtime/roles/vpcd` (placement): the role still mixes process/role supervision with network-controller responsibilities. `lease_reaper.go` (orphaned external-pool DHCP lease sweep), `ovn_controller_watchdog.go` (OVN controller liveness and restart), `host_eip_loop.go` (host-EIP reconcile loop) and `lease_datapath_reconcile.go` (lease datapath reconcile) are periodic network-control loops, not role wiring; `vpcd.go` itself also carries bridge-mode detection and verification, chassis-list parsing, external-CIDR resolution, DHCP manager startup and lease/IP-change handling, and gateway/host-EIP allocator and neighbour flush/prime construction alongside `Service.Start`. The future placement question is whether these belong in `domains/network`. Removal condition: an accepted decision on network-controller placement, and extraction of those loops and helpers behind the role.
 - ECR control-plane registration: `awsapi.Deps.NATS` is a raw `*nats.Conn`, not a narrow capability. Every NATS-backed `Handler` passes it to `domains/ecr.NewNATSMetaStore`, which calls `natsmsg.NATSRequest`, so narrowing it means changing the exported `Handler` type, every handler and the meta store; it must not be broadened further meanwhile.
 - ACM control-plane registration: `domains/acm/awsapi.Deps.NATS` (`registration.go`) is likewise a raw `*nats.Conn` carried by every `Handler`, for the same reason; it must not be broadened further meanwhile.
- `cmd/aws-model-coverage` imports `ingress/aws/dispatch`, `domains/ecr/awsapi`, `domains/acm/awsapi`, `domains/route53/awsapi` and `gateway` to read ECR's, ACM's and Route 53's declared inventories. It is a non-serving tool and the only named exception to the ADR-0002 composition invariant; it must not construct a registration.
+- `cmd/aws-model-coverage` imports `ingress/aws/dispatch`, `gateway` and the `awsapi` package of each registered service (`domains/ecr/awsapi`, `domains/acm/awsapi`, `domains/route53/awsapi`, `domains/elbv2/awsapi`, `domains/rds/awsapi`) to read their declared inventories. It is a non-serving tool and the only named exception to the ADR-0002 composition invariant; it must not construct a registration.
 - Ingress pipeline (ADR-0002 phase 4): the request-audit record, the trace action enricher, the throttle key functions and the gateway request-context keys stay in `gateway` until the auth/IAM and request-context boundary is designed. SigV4 authentication writes the audit record and context keys from four places, so moving them now would export exactly the setters ADR-0002 says must not be exported solely to make a directory move possible.
 - NAT projection delivery is inconsistent and is preserved as found: EIP `AssociateAddress` sends `vpc.add-nat` unacknowledged (`AnnounceNATAdd`), skipping the 45s OVN barrier every other add path waits on, and `DeleteNetworkInterface` sends `vpc.delete-nat` unacknowledged (`AnnounceNATDelete`), skipping the ack and the ErrNoResponders retry. The reconciler's orphan sweep is the only backstop for both. Clears when those call sites use `AddNAT`/`AddNATBestEffort` and `RemoveNAT`, a reliability change of its own.
 - `vpc.add-system-egress` has a `vpcd` subscriber and a contract entry but no producer (its former `domains/ec2/eip` publisher is gone); `domains/eks/cluster` still publishes the delete. Either a producer is missing or the add route is dead; it needs an owner decision.
@@ -456,10 +464,10 @@ These residuals were recorded by individual slices and must be resolved once the
 - `operator/installer/firstboot` (security): `Config.InstallCallback` is rendered into `spinifex-firstboot.sh`, which is world-readable (0755). The field is currently unused (`fireInstallCallback` reads `SPINIFEX_INSTALL_CALLBACK` from the environment instead), but any future wiring must not place secrets or tokenised callback URLs in the generated script; sensitive callback credentials should reach first boot through a protected channel such as a 0600 file or a credential store.
 - `operator/cli` (known security risk): `spx admin join` contacts the formation server with `InsecureSkipVerify`, sends the join token as a bearer credential and receives the cluster CA certificate and key, master key, Viperblock key and shared credentials, so an on-path attacker during formation can capture the token and the cluster root secrets or substitute a CA. This predates the move; lint surfaced it when the tree left `cmd/`. It is accepted for now and carries a line-level gosec annotation stating it is not a false positive. Removal condition: authenticate formation TLS with a pinned key or certificate delivered with the join material, or an out-of-band CA, with tests proving a mismatched server certificate fails before credentials are sent.
 - `domains/eks/cluster` (potential security risk): `ClusterMeta.ControlPlaneTemplate` (`domains/eks/cluster/service_impl.go`) clears the template's credentials but keeps `OIDCPrivateKeyPEM` and the k3s `JoinToken`, so both are persisted in the `eks-account-{acct}` cluster record, which the gateway also reads. Whether that record is encrypted at rest is unconfirmed; if it is not, a reader of the bucket can forge service-account tokens for the cluster's OIDC issuer and join nodes to it. This predates the refactor and is not fixed in it. Removal condition: the key and token are held by a dedicated secret store (a future KMS or Secrets Manager capability is the likely owner) and the cluster record keeps only a reference, with a test proving the persisted record contains neither secret.
- `operator/cli` (dependency): the command tree still imports domain and service implementations directly (`handlers/{iam,eks}`, `domains/admission/quota`, `gateway`, `domains/ochre`, `vpcd`, the Viperblock adapter and runtime roles). The ADR's `cmd/spinifex` current-to-target map row requires command packages to reach domains through public capabilities; splitting supported use cases into `operator/*` packages is later work.
+- `operator/cli` (dependency): the command tree still imports domain and service implementations directly (`handlers/{iam,eks}`, `domains/admission/quota`, `domains/ochre`, `vpcd`, the Viperblock adapter and runtime roles). The ADR's `cmd/spinifex` current-to-target map row requires command packages to reach domains through public capabilities; splitting supported use cases into `operator/*` packages is later work.
 - `providers/ebs/viperblock` → `runtime/compute/nbd` and `runtime/host/process` (dependency, moved unchanged from `services/viperblockd`): the adapter formats NBD endpoints and manages nbdkit processes through runtime packages. ADR-0001 does not list providers → runtime as forbidden and the layering test does not enforce it (same category as `providers/ebs/conformance` → `runtime/compute/nbd`), but a provider should not depend on node runtime. Clears when NBD endpoint naming has a provider-facing home (a candidate for `contracts/ebs/v1`) and process supervision is injected by the role. The adapter also imports `providers/objectstore`, a provider-to-provider edge that is allowed.
 - `gateway/ec2.go` (registration): the EC2 dispatch map stays a `GatewayConfig` method in `gateway` and imports every `domains/ec2/awsapi/*` package; `gateway/ec2_idempotency.go` also imports `domains/ec2/awsapi/idem` (`KVBucket`, `TTL`, `TokenAndParams`) for the dispatch-time idempotency wrapper. ADR-0001 makes the domain own its action registration; moving the map waits on the generic ingress registration seam and must not be done as a package move.
- `gateway` (service registration): the root `gateway` dispatch for IAM, STS, EKS, RDS, ECS and ELBv2 imports `domains/{iam,sts,eks,rds,ecs,elbv2}/awsapi`, which only AWSGW may wire. It clears when each service registers through the ADR-0002 ingress registration seam; as with EC2, moving the dispatch must not be done as a package move.
+- `gateway` (service registration): the root `gateway` dispatch for IAM, STS, EKS, RDS and ECS imports `domains/{iam,sts,eks,rds,ecs}/awsapi`, which only AWSGW may wire. It clears when each service registers through the ADR-0002 ingress registration seam; as with EC2, moving the dispatch must not be done as a package move.
 - `operator/accountlifecycle` (dependency): its EC2 reapers call `domains/ec2/awsapi/*` action adapters directly, as they did under `gateway/ec2`. The amended ADR-0001 `accountteardown` row keeps replacing these with domain-owned cleanup capabilities as deferred debt; it was not a relocation prerequisite. It also holds a read-only consumer of the EC2 platform-inventory contract (`contracts/ec2/v1`, `ec2.v1.PlatformInventory`), served by `domains/ec2/accountinventory`; only the explicit dry run (`Engine.DryRun`) calls it, to log a shadow comparison, and the key-pair, placement-group and launch-template reapers still use the `awsapi` adapters. A domain-owned cleanup capability for the proven-safe subset of those three kinds (key pairs and empty placement groups no instance, spot request or live launch-template version names, and launch-template residue; never live launch templates) exists as `domains/ec2/accountcleanup` (a Go interface with injected owners and reference readers, no NATS subject); nothing calls it yet. It also holds the StageAccount settings-seal reaper, a consumer of `contracts/ec2/v1`'s `ec2.v1.SealAccountSettings` and per-node `ec2.v1.AccountSettingsSealCapability.<node>` subjects, served by `domains/ec2/account` through the daemon; the reaper reads no KV and imports no domain package, reaching both subjects only through NATS and the configured node set `NewClusterEngine` already receives.
 - `domains/admission/quota` (dependency, moved from `handlers/quota` unchanged): `live.go` imports `domains/elbv2/awsapi` and `domains/rds/instance` for its live ELBv2 and RDS counters; `records.go` imports `runtime/compute/vm` for the instance-record vCPU sweep; `live.go`, `vcpu.go` and `storage.go` call `domains/ec2/awsapi/{vpc,eip,volume}` describe adapters to count VPCs, subnets, addresses and volumes, as they did under `gateway/ec2`. The ADR's `handlers/quota` current-to-target map row replaces these with reader capabilities once their final locations exist.
 - `domains/ec2/awsapi/instance` (dependency): imports `domains/iam` (`IAMService` in `DescribeInstances.go`, `RunInstances.go` and `IamInstanceProfileAssociation.go`; `InstanceProfile` in the latter two), `runtime/compute/cache` (`NodeState`, `NodeUnknown`, `NodeLive`) and `runtime/compute/vm` (`VM`), both in `status_synthesis.go`. The IAM edge needs an IAM capability; the runtime edges are domain-to-runtime and need the instance record and node liveness exposed through an EC2-owned projection.
@@ -676,8 +684,8 @@ series above; "needs decision" is open.
 | `gateway/bedrock` | — | Done. The whole package moved into `domains/ochre` in `spinifex#1266` with its package-private core and tests intact and no behaviour change. Separating the action adapters into `domains/ochre/awsapi` remains deferred, recorded as `domains/ochre` (action adapters) debt above. |
 | `gateway/{iam,sts,eks,rds}` | — | Done. The action adapters moved to `domains/{iam,sts,eks,rds}/awsapi` in `spinifex#1265` with no API, NATS subject, storage or behaviour change. ADR-0001 does not make ADR-0002 registration a prerequisite for relocating a domain's adapters, as the EC2 adapter moves already showed; the root `gateway` dispatch that still imports them is recorded `gateway` (service registration) debt. |
 | `gateway/{ecs,elbv2}` | — | Done. The action adapters moved to `domains/{ecs,elbv2}/awsapi` in `spinifex#1273` with no API, NATS subject, storage or behaviour change; the root `gateway` dispatch that still imports them is recorded `gateway` (service registration) debt. |
-| `gateway/spx` | `promote_image.go` and `spx.go` dispatch (`storage.go` already consumes `contracts/operator/v1`). | Needs decision: which package owns operator-plane action dispatch registration — an `ingress/aws`-style seam for the operator plane, or `gateway` stays its home. |
-| `gateway/tagging` | — | Done. `domains/tagging` owns the read-only cross-resource `GetResources` aggregation and its API semantics (`0d830e901`, `aedd0b2c4`); AWSGW composes the owner listers, the EC2 tag index and ELBv2 (`f3545e3ea`); the root `gateway` keeps only the legacy dispatch in `gateway/tagging.go`. The package is removed, and its two allowlist rows and frozen-legacy entry are gone. Owner tag mutation capabilities remain future direction, so the RDS and ECS tag-writing edges above stay open. |
+| `gateway/spx` | — | Done. The gather helpers, storage-status config and image-promote request moved to `operator/admin` with no behaviour change and the package is removed. The `spinifex` Query dispatch and the `/admin/{method}` handlers that call them now live in `operator/admin` as one seam registration. |
+| `gateway/tagging` | — | Done. `domains/tagging` owns the read-only cross-resource `GetResources` aggregation and its API semantics (`0d830e901`, `aedd0b2c4`); AWSGW composes the owner listers, the EC2 tag index and ELBv2 (`f3545e3ea`); the dispatch moved to `domains/tagging/awsapi` and the root `gateway` no longer serves Tagging. The package is removed, and its two allowlist rows and frozen-legacy entry are gone. Owner tag mutation capabilities remain future direction, so the RDS and ECS tag-writing edges above stay open. |
 | `services/viperblockd` | — | Done. The adapter is `providers/ebs/viperblock` (with the read-only `dirty` reader), the shared vocabulary is `contracts/viperblockd/legacy/v1`, and the role is `runtime/roles/viperblockd`. `services/` is empty. |
 | `utils` | — | Done and deleted. `ec2_subject.go` went to `contracts/ec2/v1`, the vpcd protocol to `contracts/network/v1` with the typed client in `domains/network/projection`, and the master-key loader to `domains/ec2/ebs/encryption`. |
 | `vpcd` | — | Done. The IMDS local-file reader and record-space opener moved to `domains/ec2/instance/state` and are injected through `guestmetadata.LocalStateReader`/`RecordLoader` (recorded above), so `vpcd` no longer imports `daemon` directly; its inter-process boundary now exists (`contracts/network/v1`, with `domains/network/subscribers` as the consumer and `domains/network/projection` as the producer client). The package moved unchanged to `runtime/roles/vpcd` (recorded below), and the legacy directory is removed. |
@@ -696,7 +704,7 @@ ADR-0001 completion is structural, so none of these is cleared by remediating se
 | Class | Paths | Status |
 |---|---|---|
 | No missing boundary (candidate moves) | `handlers/ecs/bus` → `contracts/ecs/v1` (`a130f6b8e`); `admin/images_promote.go` → `domains/ec2/image` (`263c58167`); `admin` bootstrap account identifiers → `foundation/aws/identifiers` (`3fdaa54d9`); `admin` credential-format generators → `foundation/aws/credentials` `0368e861a`, retiring `admin` | Done. |
-| One narrow missing boundary | Done: `handlers/sts` EKS JWKS lookup → `handlers/eks` `ClusterJWKSResolver` (`254d898a3`); `gateway/tagging` → `domains/tagging` with AWSGW-composed listers (`0d830e901`, `f3545e3ea`, `aedd0b2c4`). Open: `gateway/spx` (owner of operator-plane dispatch registration) | Each needs one decision or one capability, then a mechanical move. |
+| One narrow missing boundary | Done: `handlers/sts` EKS JWKS lookup → `handlers/eks` `ClusterJWKSResolver` (`254d898a3`); `gateway/tagging` → `domains/tagging` with AWSGW-composed listers (`0d830e901`, `f3545e3ea`, `aedd0b2c4`); `gateway/spx` → `operator/admin`. Done: the `spinifex` service and `/admin` dispatch registration, owned by `operator/admin` | Each needs one decision or one capability, then a mechanical move. |
 | Broader boundary or follows another package | `gateway` root (ADR-0002 auth and request-context phase) | Recorded debt with removal conditions above. |

 Per-consumer IAM capabilities are the widest unblocker: they gate STS's IAM reads, Ochre's IAM reads, the ELBv2 role path and the ECS, EKS and RDS ensurers.
@@ -73,7 +73,6 @@ The cluster plane is peer-scoped. On a single-NIC node the planes collapse onto
 | Ports | Serving |
 |---|---|
 | TCP 4222, 4248 | NATS client and cluster routes |
-| TCP 4223, 4249 | SQS message store client and cluster routes |
 | TCP 4432 | Formation, then the daemon cluster manager |
 | TCP and UDP 5300 | Internal DNS |
 | TCP 6641, 6642 | OVN northbound and southbound |
@@ -158,6 +157,7 @@ Restarting the daemon is what re-arms it: the node rebuilds its peer list from t

 > [!WARNING]
 > Do not flush the whole nftables ruleset to clear the policy. The VPC daemon writes SNAT and per-elastic-IP forward rules into the `ip nat` and `ip filter` tables and reinstalls them only when it starts, so `nft flush ruleset` breaks elastic IPs silently until the next `systemctl restart spinifex-vpcd`. Use `disable`, which only ever touches `inet spinifex_filter`.
+> On Oracle Linux with IPsec on, a flush also removes `inet spinifex_ipsec_guard`, so Geneve to a peer without an SA goes out in clear until the daemon puts the table back, within 60 seconds.

 ## The Files

@@ -2,8 +2,8 @@
  "schema_version": 1,
  "model_source": {
    "repository": "github.com/aws/api-models-aws",
-    "commit": "9b1a4929274443d8671ea8acfda7afbe856e29e9",
-    "commit_date": "2026-09-25"
+    "commit": "e10d6a1c5ac36af7c0c61d1f67b3c3ebbe41de49",
+    "commit_date": "2026-10-28"
  },
  "services": [
    {
@@ -300,6 +300,7 @@
          "DeleteCapacityManagerDataExport",
          "DeleteCarrierGateway",
          "DeleteClientVpnEndpoint",
+          "DeleteClientVpnEndpointAuthorizationPolicy",
          "DeleteClientVpnRoute",
          "DeleteCoipCidr",
          "DeleteCoipPool",
@@ -681,7 +682,9 @@
          "GetCapacityManagerMetricData",
          "GetCapacityManagerMetricDimensions",
          "GetCapacityManagerMonitoredTagKeys",
+          "GetCapacityManagerSupportabilityScores",
          "GetCapacityReservationUsage",
+          "GetClientVpnEndpointAuthorizationPolicy",
          "GetCoipPoolUsage",
          "GetConsoleOutput",
          "GetConsoleScreenshot",
@@ -695,6 +698,7 @@
          "GetHostReservationPurchasePreview",
          "GetImageAncestry",
          "GetImageBlockPublicAccessState",
+          "GetImageFindings",
          "GetInstanceMetadataDefaults",
          "GetInstanceTpmEkPub",
          "GetInstanceTypesFromInstanceRequirements",
@@ -734,6 +738,7 @@
          "GetSnapshotBlockPublicAccessState",
          "GetSpotPlacementScores",
          "GetSubnetCidrReservations",
+          "GetTransitGatewayAttachmentNatConfiguration",
          "GetTransitGatewayAttachmentPropagations",
          "GetTransitGatewayMeteringPolicyEntries",
          "GetTransitGatewayMulticastDomainAssociations",
@@ -766,6 +771,7 @@
          "ModifyCapacityReservation",
          "ModifyCapacityReservationFleet",
          "ModifyClientVpnEndpoint",
+          "ModifyClientVpnEndpointAuthorizationPolicy",
          "ModifyDefaultCreditSpecification",
          "ModifyEbsDefaultKmsKeyId",
          "ModifyFleet",
@@ -814,6 +820,7 @@
          "ModifyTrafficMirrorFilterRule",
          "ModifyTrafficMirrorSession",
          "ModifyTransitGateway",
+          "ModifyTransitGatewayAttachmentNatConfiguration",
          "ModifyTransitGatewayMeteringPolicy",
          "ModifyTransitGatewayPolicyTableEntry",
          "ModifyTransitGatewayPrefixListReference",
@@ -1335,6 +1342,7 @@
          "DeleteCapacityManagerDataExport",
          "DeleteCarrierGateway",
          "DeleteClientVpnEndpoint",
+          "DeleteClientVpnEndpointAuthorizationPolicy",
          "DeleteClientVpnRoute",
          "DeleteCoipCidr",
          "DeleteCoipPool",
@@ -1653,7 +1661,9 @@
          "GetCapacityManagerMetricData",
          "GetCapacityManagerMetricDimensions",
          "GetCapacityManagerMonitoredTagKeys",
+          "GetCapacityManagerSupportabilityScores",
          "GetCapacityReservationUsage",
+          "GetClientVpnEndpointAuthorizationPolicy",
          "GetCoipPoolUsage",
          "GetConsoleScreenshot",
          "GetDeclarativePoliciesReportSummary",
@@ -1665,6 +1675,7 @@
          "GetHostReservationPurchasePreview",
          "GetImageAncestry",
          "GetImageBlockPublicAccessState",
+          "GetImageFindings",
          "GetInstanceMetadataDefaults",
          "GetInstanceTpmEkPub",
          "GetInstanceTypesFromInstanceRequirements",
@@ -1701,6 +1712,7 @@
          "GetSnapshotBlockPublicAccessState",
          "GetSpotPlacementScores",
          "GetSubnetCidrReservations",
+          "GetTransitGatewayAttachmentNatConfiguration",
          "GetTransitGatewayAttachmentPropagations",
          "GetTransitGatewayMeteringPolicyEntries",
          "GetTransitGatewayMulticastDomainAssociations",
@@ -1732,6 +1744,7 @@
          "ModifyCapacityReservation",
          "ModifyCapacityReservationFleet",
          "ModifyClientVpnEndpoint",
+          "ModifyClientVpnEndpointAuthorizationPolicy",
          "ModifyDefaultCreditSpecification",
          "ModifyEbsDefaultKmsKeyId",
          "ModifyFleet",
@@ -1773,6 +1786,7 @@
          "ModifyTrafficMirrorFilterRule",
          "ModifyTrafficMirrorSession",
          "ModifyTransitGateway",
+          "ModifyTransitGatewayAttachmentNatConfiguration",
          "ModifyTransitGatewayMeteringPolicy",
          "ModifyTransitGatewayPolicyTableEntry",
          "ModifyTransitGatewayPrefixListReference",
@@ -1897,6 +1911,7 @@
          "DescribeImages",
          "DescribePullThroughCacheRules",
          "DescribeRegistry",
+          "DescribeReplicationBackfillResults",
          "DescribeRepositories",
          "DescribeRepositoryCreationTemplates",
          "GetAccountSetting",
@@ -1912,6 +1927,7 @@
          "ListImageReferrers",
          "ListImages",
          "ListPullTimeUpdateExclusions",
+          "ListReplicationBackfills",
          "ListTagsForResource",
          "PutAccountSetting",
          "PutImage",
@@ -1926,6 +1942,8 @@
          "SetRepositoryPolicy",
          "StartImageScan",
          "StartLifecyclePolicyPreview",
+          "StartReplicationBackfill",
+          "StopReplicationBackfill",
          "TagResource",
          "UntagResource",
          "UpdateImageStorageClass",
@@ -2031,14 +2049,18 @@
          "DescribeImageReplicationStatus",
          "DescribeImageSigningStatus",
          "DescribePullThroughCacheRules",
+          "DescribeReplicationBackfillResults",
          "DescribeRepositoryCreationTemplates",
          "GetAccountSetting",
          "GetSigningConfiguration",
          "ListImageReferrers",
          "ListPullTimeUpdateExclusions",
+          "ListReplicationBackfills",
          "PutAccountSetting",
          "PutSigningConfiguration",
          "RegisterPullTimeUpdateExclusion",
+          "StartReplicationBackfill",
+          "StopReplicationBackfill",
          "UpdateImageStorageClass",
          "UpdatePullThroughCacheRule",
          "UpdateRepositoryCreationTemplate",
@@ -2058,11 +2080,13 @@
      "note": "",
      "operations": {
        "modelled": [
+          "AssociateGatewayWithCluster",
          "ContinueServiceDeployment",
          "CreateCapacityProvider",
          "CreateCluster",
          "CreateDaemon",
          "CreateExpressGatewayService",
+          "CreateGateway",
          "CreateService",
          "CreateTaskSet",
          "DeleteAccountSetting",
@@ -2072,6 +2096,7 @@
          "DeleteDaemon",
          "DeleteDaemonTaskDefinition",
          "DeleteExpressGatewayService",
+          "DeleteGateway",
          "DeleteService",
          "DeleteTaskDefinitions",
          "DeleteTaskSet",
@@ -2085,12 +2110,16 @@
          "DescribeDaemonRevisions",
          "DescribeDaemonTaskDefinition",
          "DescribeExpressGatewayService",
+          "DescribeGateway",
+          "DescribeGatewayResources",
+          "DescribeGatewayRoute",
          "DescribeServiceDeployments",
          "DescribeServiceRevisions",
          "DescribeServices",
          "DescribeTaskDefinition",
          "DescribeTaskSets",
          "DescribeTasks",
+          "DisassociateGatewayFromCluster",
          "DiscoverPollEndpoint",
          "ExecuteCommand",
          "GetTaskProtection",
@@ -2101,6 +2130,9 @@
          "ListDaemonDeployments",
          "ListDaemonTaskDefinitions",
          "ListDaemons",
+          "ListGatewayClusterAssociations",
+          "ListGatewayRoutes",
+          "ListGateways",
          "ListServiceDeployments",
          "ListServices",
          "ListServicesByNamespace",
@@ -2131,6 +2163,7 @@
          "UpdateContainerInstancesState",
          "UpdateDaemon",
          "UpdateExpressGatewayService",
+          "UpdateGateway",
          "UpdateService",
          "UpdateServicePrimaryTaskSet",
          "UpdateTaskProtection",
@@ -2219,15 +2252,18 @@
        ],
        "unsupported": [],
        "missing": [
+          "AssociateGatewayWithCluster",
          "ContinueServiceDeployment",
          "CreateDaemon",
          "CreateExpressGatewayService",
+          "CreateGateway",
          "CreateTaskSet",
          "DeleteAccountSetting",
          "DeleteAttributes",
          "DeleteDaemon",
          "DeleteDaemonTaskDefinition",
          "DeleteExpressGatewayService",
+          "DeleteGateway",
          "DeleteTaskDefinitions",
          "DeleteTaskSet",
          "DescribeDaemon",
@@ -2235,9 +2271,13 @@
          "DescribeDaemonRevisions",
          "DescribeDaemonTaskDefinition",
          "DescribeExpressGatewayService",
+          "DescribeGateway",
+          "DescribeGatewayResources",
+          "DescribeGatewayRoute",
          "DescribeServiceDeployments",
          "DescribeServiceRevisions",
          "DescribeTaskSets",
+          "DisassociateGatewayFromCluster",
          "DiscoverPollEndpoint",
          "ExecuteCommand",
          "GetTaskProtection",
@@ -2245,6 +2285,9 @@
          "ListDaemonDeployments",
          "ListDaemonTaskDefinitions",
          "ListDaemons",
+          "ListGatewayClusterAssociations",
+          "ListGatewayRoutes",
+          "ListGateways",
          "ListServiceDeployments",
          "PutAccountSettingDefault",
          "PutAttributes",
@@ -2257,6 +2300,7 @@
          "UpdateContainerAgent",
          "UpdateDaemon",
          "UpdateExpressGatewayService",
+          "UpdateGateway",
          "UpdateServicePrimaryTaskSet",
          "UpdateTaskProtection",
          "UpdateTaskSet"
@@ -3099,6 +3143,7 @@
          "ApplyPendingMaintenanceAction",
          "AuthorizeDBSecurityGroupIngress",
          "BacktrackDBCluster",
+          "CancelDBCapacityReservation",
          "CancelExportTask",
          "CopyDBClusterParameterGroup",
          "CopyDBClusterSnapshot",
@@ -3107,6 +3152,7 @@
          "CopyOptionGroup",
          "CreateBlueGreenDeployment",
          "CreateCustomDBEngineVersion",
+          "CreateDBCapacityReservation",
          "CreateDBCluster",
          "CreateDBClusterEndpoint",
          "CreateDBClusterParameterGroup",
@@ -3150,6 +3196,7 @@
          "DescribeAccountAttributes",
          "DescribeBlueGreenDeployments",
          "DescribeCertificates",
+          "DescribeDBCapacityReservations",
          "DescribeDBClusterAutomatedBackups",
          "DescribeDBClusterBacktracks",
          "DescribeDBClusterEndpoints",
@@ -3204,6 +3251,7 @@
          "ModifyCertificates",
          "ModifyCurrentDBClusterCapacity",
          "ModifyCustomDBEngineVersion",
+          "ModifyDBCapacityReservation",
          "ModifyDBCluster",
          "ModifyDBClusterEndpoint",
          "ModifyDBClusterParameterGroup",
@@ -3354,6 +3402,7 @@
          "ApplyPendingMaintenanceAction",
          "AuthorizeDBSecurityGroupIngress",
          "BacktrackDBCluster",
+          "CancelDBCapacityReservation",
          "CancelExportTask",
          "CopyDBClusterParameterGroup",
          "CopyDBClusterSnapshot",
@@ -3362,6 +3411,7 @@
          "CopyOptionGroup",
          "CreateBlueGreenDeployment",
          "CreateCustomDBEngineVersion",
+          "CreateDBCapacityReservation",
          "CreateDBClusterEndpoint",
          "CreateDBClusterParameterGroup",
          "CreateDBClusterSnapshot",
@@ -3392,6 +3442,7 @@
          "DescribeAccountAttributes",
          "DescribeBlueGreenDeployments",
          "DescribeCertificates",
+          "DescribeDBCapacityReservations",
          "DescribeDBClusterAutomatedBackups",
          "DescribeDBClusterBacktracks",
          "DescribeDBClusterEndpoints",
@@ -3433,6 +3484,7 @@
          "ModifyCertificates",
          "ModifyCurrentDBClusterCapacity",
          "ModifyCustomDBEngineVersion",
+          "ModifyDBCapacityReservation",
          "ModifyDBClusterEndpoint",
          "ModifyDBClusterParameterGroup",
          "ModifyDBClusterSnapshotAttribute",
@@ -28,6 +28,21 @@ Spot Instance Requests are a mock over the on-demand `RunInstances` path. A requ

 An AMI cannot be shared with another account or made public. `ModifyImageAttribute` refuses `LaunchPermission`, `UserIds`, `UserGroups`, `OrganizationArns`, `OrganizationalUnitArns` and `OperationType`, and `ResetImageAttribute` resets only `description`, not `launchPermission`. Each refusal is `InvalidParameterValue`. Terraform's `aws_ami_launch_permission` therefore cannot be created. `description` is the only image attribute that can be modified or reset.

+### RegisterImage Registers From a Snapshot
+
+`RegisterImage` takes a root EBS snapshot in `BlockDeviceMappings` and writes an image that points at it.
+`EnaSupport` is supported, with `true` and `false` both accepted and stored on the image.
+`DescribeImages` returns `EnaSupport` when the image has one and leaves the member out when it was never set, as AWS does.
+`CopyImage` carries the value to the copy, and `CreateImage` sets it to `true` because every instance reports ENA support.
+`RunInstances` does not read the flag, so an image registered with `EnaSupport=false` is not held back from instance types that need ENA on AWS, since guests use virtio networking either way.
+
+`RegisterImage` refuses `ImageLocation`, `KernelId`, `RamdiskId`, `TpmSupport`, `ImdsSupport` and `SriovNetSupport`, and any `VirtualizationType` other than `hvm`.
+Each refusal is `InvalidParameterValue`, and the message names the parameter and the value, for example `Value (aki-123) for parameter KernelId is invalid. KernelId is not supported.`
+The same form is used for a bad `Architecture`, `BootMode` or `Name` length.
+A missing `Name`, `BlockDeviceMappings` or root snapshot answers `MissingParameter` and names what is missing.
+Registering with `--ena-support`, `--no-ena-support` and neither, copying the image, and three refusals were run on a three-node cluster with the AWS CLI.
+Launching an instance from an `EnaSupport` image was not run, so the launch path is unchanged by construction and not by test.
+
 ### Operations

 | Operation |
@@ -5,3 +5,18 @@ Spot Instance Requests are a mock over the on-demand `RunInstances` path. A requ
 ### AMI Launch Permissions Are Refused

 An AMI cannot be shared with another account or made public. `ModifyImageAttribute` refuses `LaunchPermission`, `UserIds`, `UserGroups`, `OrganizationArns`, `OrganizationalUnitArns` and `OperationType`, and `ResetImageAttribute` resets only `description`, not `launchPermission`. Each refusal is `InvalidParameterValue`. Terraform's `aws_ami_launch_permission` therefore cannot be created. `description` is the only image attribute that can be modified or reset.
+
+### RegisterImage Registers From a Snapshot
+
+`RegisterImage` takes a root EBS snapshot in `BlockDeviceMappings` and writes an image that points at it.
+`EnaSupport` is supported, with `true` and `false` both accepted and stored on the image.
+`DescribeImages` returns `EnaSupport` when the image has one and leaves the member out when it was never set, as AWS does.
+`CopyImage` carries the value to the copy, and `CreateImage` sets it to `true` because every instance reports ENA support.
+`RunInstances` does not read the flag, so an image registered with `EnaSupport=false` is not held back from instance types that need ENA on AWS, since guests use virtio networking either way.
+
+`RegisterImage` refuses `ImageLocation`, `KernelId`, `RamdiskId`, `TpmSupport`, `ImdsSupport` and `SriovNetSupport`, and any `VirtualizationType` other than `hvm`.
+Each refusal is `InvalidParameterValue`, and the message names the parameter and the value, for example `Value (aki-123) for parameter KernelId is invalid. KernelId is not supported.`
+The same form is used for a bad `Architecture`, `BootMode` or `Name` length.
+A missing `Name`, `BlockDeviceMappings` or root snapshot answers `MissingParameter` and names what is missing.
+Registering with `--ena-support`, `--no-ena-support` and neither, copying the image, and three refusals were run on a three-node cluster with the AWS CLI.
+Launching an instance from an `EnaSupport` image was not run, so the launch path is unchanged by construction and not by test.
@@ -52,6 +52,46 @@ Measured on a three-node cluster with a 5 second interval and thresholds of 2:
 A TCP connection held through a Network Load Balancer survived four configuration reloads while the other backend changed state, because the agent applies nginx with a graceful reload and HAProxy with a graceful restart.
 The HAProxy configuration with the backup servers was accepted by a live Application Load Balancer, but its rendered text was not dumped and checked by hand.

+### Forwarding headers
+
+An Application Load Balancer sets three headers on every HTTP and HTTPS request it forwards.
+`X-Forwarded-For` carries the client address, `X-Forwarded-Proto` is `http` or `https` for the listener that accepted the request, and `X-Forwarded-Port` is that listener's port.
+`X-Forwarded-Proto` and `X-Forwarded-Port` are always overwritten, so a value sent by the client never reaches the target.
+Network Load Balancers do not touch the stream.
+
+`routing.http.xff_header_processing.mode` decides what happens to `X-Forwarded-For`:
+
+| Mode | Behaviour |
+| --- | --- |
+| `append` (default) | Keeps what the client sent and adds the client address last, as one comma-separated line such as `203.0.113.9, 192.0.2.10`. |
+| `preserve` | Forwards the client's value unchanged and adds nothing. The header is absent when the client sent none. |
+| `remove` | Drops the header, including any client value. |
+
+`routing.http.xff_client_port.enabled` (default `false`) appends the client's source port in `append` mode, as `192.0.2.10:54321`.
+An IPv6 client is written `[2001:db8::1]:54321`.
+The setting has no effect in `preserve` and `remove` modes.
+
+`ModifyLoadBalancerAttributes` accepts only `append`, `preserve` or `remove` for the mode and `true` or `false` for the client port, and rejects anything else with `ValidationError`.
+A change re-renders the load balancer configuration, and a call that leaves both values as they were does not.
+In a live test each change reached the target within one 3 second poll.
+
+A client that sends `X-Forwarded-For` more than once has its lines joined into the single comma-separated value.
+At most eight client lines are joined, and any beyond the eighth are dropped, while the client address is still added last.
+
+Every other `routing.http.*` attribute is stored and returned by `DescribeLoadBalancerAttributes` but has no effect on the data plane.
+
+Measured on a three-node cluster, with the caller on the first node:
+
+| Listener | `X-Forwarded-For` the target saw |
+| --- | --- |
+| Internet-facing HTTP | The node's real address, `192.168.0.120` |
+| Internet-facing HTTPS | The same address, with `https` and port 443 |
+| Internal HTTP | The calling instance's private address |
+| Internet-facing HTTP with a client value of `203.0.113.9` | `203.0.113.9, 192.168.0.120` on one line |
+
+A packet capture on the load balancer's interface showed HAProxy receiving the real client address, so nothing translates it first.
+Behaviour against AWS was not compared, and the cases above were not run with an IPv6 client.
+
 ### Operations

 | Operation |
@@ -29,3 +29,43 @@ Measured on a three-node cluster with a 5 second interval and thresholds of 2:

 A TCP connection held through a Network Load Balancer survived four configuration reloads while the other backend changed state, because the agent applies nginx with a graceful reload and HAProxy with a graceful restart.
 The HAProxy configuration with the backup servers was accepted by a live Application Load Balancer, but its rendered text was not dumped and checked by hand.
+
+### Forwarding headers
+
+An Application Load Balancer sets three headers on every HTTP and HTTPS request it forwards.
+`X-Forwarded-For` carries the client address, `X-Forwarded-Proto` is `http` or `https` for the listener that accepted the request, and `X-Forwarded-Port` is that listener's port.
+`X-Forwarded-Proto` and `X-Forwarded-Port` are always overwritten, so a value sent by the client never reaches the target.
+Network Load Balancers do not touch the stream.
+
+`routing.http.xff_header_processing.mode` decides what happens to `X-Forwarded-For`:
+
+| Mode | Behaviour |
+| --- | --- |
+| `append` (default) | Keeps what the client sent and adds the client address last, as one comma-separated line such as `203.0.113.9, 192.0.2.10`. |
+| `preserve` | Forwards the client's value unchanged and adds nothing. The header is absent when the client sent none. |
+| `remove` | Drops the header, including any client value. |
+
+`routing.http.xff_client_port.enabled` (default `false`) appends the client's source port in `append` mode, as `192.0.2.10:54321`.
+An IPv6 client is written `[2001:db8::1]:54321`.
+The setting has no effect in `preserve` and `remove` modes.
+
+`ModifyLoadBalancerAttributes` accepts only `append`, `preserve` or `remove` for the mode and `true` or `false` for the client port, and rejects anything else with `ValidationError`.
+A change re-renders the load balancer configuration, and a call that leaves both values as they were does not.
+In a live test each change reached the target within one 3 second poll.
+
+A client that sends `X-Forwarded-For` more than once has its lines joined into the single comma-separated value.
+At most eight client lines are joined, and any beyond the eighth are dropped, while the client address is still added last.
+
+Every other `routing.http.*` attribute is stored and returned by `DescribeLoadBalancerAttributes` but has no effect on the data plane.
+
+Measured on a three-node cluster, with the caller on the first node:
+
+| Listener | `X-Forwarded-For` the target saw |
+| --- | --- |
+| Internet-facing HTTP | The node's real address, `192.168.0.120` |
+| Internet-facing HTTPS | The same address, with `https` and port 443 |
+| Internal HTTP | The calling instance's private address |
+| Internet-facing HTTP with a client value of `203.0.113.9` | `203.0.113.9, 192.168.0.120` on one line |
+
+A packet capture on the load balancer's interface showed HAProxy receiving the real client address, so nothing translates it first.
+Behaviour against AWS was not compared, and the cases above were not run with an IPv6 client.
@@ -4,7 +4,7 @@ Generated by `make aws-model-coverage`. This file is not published: it carries t

 **Not implemented** is the candidate work. **Not applicable** is declared in `pages.json` and is not work: those operations describe something this platform will not have.

-Pinned model source: `github.com/aws/api-models-aws@9b1a49292744, 2026-09-25`.
+Pinned model source: `github.com/aws/api-models-aws@e10d6a1c5ac3, 2026-10-28`.

 ## acm

@@ -46,7 +46,7 @@ Pinned model source: `github.com/aws/api-models-aws@9b1a49292744, 2026-09-25`.

 ## ec2

-`2016-11-15` — 127 of 804 modelled operations implemented (15.8%), 0 stubbed, 39 not applicable, 0 registered outside the pinned model.
+`2016-11-15` — 127 of 811 modelled operations implemented (15.7%), 0 stubbed, 39 not applicable, 0 registered outside the pinned model.

 ### Not implemented

@@ -190,6 +190,7 @@ Pinned model source: `github.com/aws/api-models-aws@9b1a49292744, 2026-09-25`.
 - `DeleteApplicationStatusCheck`
 - `DeleteCapacityManagerDataExport`
 - `DeleteClientVpnEndpoint`
+- `DeleteClientVpnEndpointAuthorizationPolicy`
 - `DeleteClientVpnRoute`
 - `DeleteCoipCidr`
 - `DeleteCoipPool`
@@ -484,7 +485,9 @@ Pinned model source: `github.com/aws/api-models-aws@9b1a49292744, 2026-09-25`.
 - `GetCapacityManagerMetricData`
 - `GetCapacityManagerMetricDimensions`
 - `GetCapacityManagerMonitoredTagKeys`
+- `GetCapacityManagerSupportabilityScores`
 - `GetCapacityReservationUsage`
+- `GetClientVpnEndpointAuthorizationPolicy`
 - `GetCoipPoolUsage`
 - `GetConsoleScreenshot`
 - `GetDeclarativePoliciesReportSummary`
@@ -496,6 +499,7 @@ Pinned model source: `github.com/aws/api-models-aws@9b1a49292744, 2026-09-25`.
 - `GetHostReservationPurchasePreview`
 - `GetImageAncestry`
 - `GetImageBlockPublicAccessState`
+- `GetImageFindings`
 - `GetInstanceMetadataDefaults`
 - `GetInstanceTpmEkPub`
 - `GetInstanceTypesFromInstanceRequirements`
@@ -532,6 +536,7 @@ Pinned model source: `github.com/aws/api-models-aws@9b1a49292744, 2026-09-25`.
 - `GetSnapshotBlockPublicAccessState`
 - `GetSpotPlacementScores`
 - `GetSubnetCidrReservations`
+- `GetTransitGatewayAttachmentNatConfiguration`
 - `GetTransitGatewayAttachmentPropagations`
 - `GetTransitGatewayMeteringPolicyEntries`
 - `GetTransitGatewayMulticastDomainAssociations`
@@ -563,6 +568,7 @@ Pinned model source: `github.com/aws/api-models-aws@9b1a49292744, 2026-09-25`.
 - `ModifyCapacityReservation`
 - `ModifyCapacityReservationFleet`
 - `ModifyClientVpnEndpoint`
+- `ModifyClientVpnEndpointAuthorizationPolicy`
 - `ModifyDefaultCreditSpecification`
 - `ModifyEbsDefaultKmsKeyId`
 - `ModifyFleet`
@@ -601,6 +607,7 @@ Pinned model source: `github.com/aws/api-models-aws@9b1a49292744, 2026-09-25`.
 - `ModifyTrafficMirrorFilterRule`
 - `ModifyTrafficMirrorSession`
 - `ModifyTransitGateway`
+- `ModifyTransitGatewayAttachmentNatConfiguration`
 - `ModifyTransitGatewayMeteringPolicy`
 - `ModifyTransitGatewayPolicyTableEntry`
 - `ModifyTransitGatewayPrefixListReference`
@@ -762,7 +769,7 @@ spot-price: On owned hardware there is no spot-to-on-demand price differential,

 ## ecr

-`2015-09-21` — 23 of 58 modelled operations implemented (39.7%), 10 stubbed, 0 not applicable, 3 registered outside the pinned model.
+`2015-09-21` — 23 of 62 modelled operations implemented (37.1%), 10 stubbed, 0 not applicable, 3 registered outside the pinned model.

 ### Not implemented

@@ -776,14 +783,18 @@ spot-price: On owned hardware there is no spot-to-on-demand price differential,
 - `DescribeImageReplicationStatus`
 - `DescribeImageSigningStatus`
 - `DescribePullThroughCacheRules`
+- `DescribeReplicationBackfillResults`
 - `DescribeRepositoryCreationTemplates`
 - `GetAccountSetting`
 - `GetSigningConfiguration`
 - `ListImageReferrers`
 - `ListPullTimeUpdateExclusions`
+- `ListReplicationBackfills`
 - `PutAccountSetting`
 - `PutSigningConfiguration`
 - `RegisterPullTimeUpdateExclusion`
+- `StartReplicationBackfill`
+- `StopReplicationBackfill`
 - `UpdateImageStorageClass`
 - `UpdatePullThroughCacheRule`
 - `UpdateRepositoryCreationTemplate`
@@ -819,19 +830,22 @@ spot-price: On owned hardware there is no spot-to-on-demand price differential,

 ## ecs

-`2014-11-13` — 31 of 77 modelled operations implemented (40.3%), 5 stubbed, 0 not applicable, 3 registered outside the pinned model.
+`2014-11-13` — 31 of 88 modelled operations implemented (35.2%), 5 stubbed, 0 not applicable, 3 registered outside the pinned model.

 ### Not implemented

+- `AssociateGatewayWithCluster`
 - `ContinueServiceDeployment`
 - `CreateDaemon`
 - `CreateExpressGatewayService`
+- `CreateGateway`
 - `CreateTaskSet`
 - `DeleteAccountSetting`
 - `DeleteAttributes`
 - `DeleteDaemon`
 - `DeleteDaemonTaskDefinition`
 - `DeleteExpressGatewayService`
+- `DeleteGateway`
 - `DeleteTaskDefinitions`
 - `DeleteTaskSet`
 - `DescribeDaemon`
@@ -839,9 +853,13 @@ spot-price: On owned hardware there is no spot-to-on-demand price differential,
 - `DescribeDaemonRevisions`
 - `DescribeDaemonTaskDefinition`
 - `DescribeExpressGatewayService`
+- `DescribeGateway`
+- `DescribeGatewayResources`
+- `DescribeGatewayRoute`
 - `DescribeServiceDeployments`
 - `DescribeServiceRevisions`
 - `DescribeTaskSets`
+- `DisassociateGatewayFromCluster`
 - `DiscoverPollEndpoint`
 - `ExecuteCommand`
 - `GetTaskProtection`
@@ -849,6 +867,9 @@ spot-price: On owned hardware there is no spot-to-on-demand price differential,
 - `ListDaemonDeployments`
 - `ListDaemonTaskDefinitions`
 - `ListDaemons`
+- `ListGatewayClusterAssociations`
+- `ListGatewayRoutes`
+- `ListGateways`
 - `ListServiceDeployments`
 - `PutAccountSettingDefault`
 - `PutAttributes`
@@ -861,6 +882,7 @@ spot-price: On owned hardware there is no spot-to-on-demand price differential,
 - `UpdateContainerAgent`
 - `UpdateDaemon`
 - `UpdateExpressGatewayService`
+- `UpdateGateway`
 - `UpdateServicePrimaryTaskSet`
 - `UpdateTaskProtection`
 - `UpdateTaskSet`
@@ -1085,7 +1107,7 @@ signing-certificates: X.509 signing certificates are an EC2-Classic SOAP credent

 ## rds

-`2014-10-31` — 27 of 164 modelled operations implemented (16.5%), 0 stubbed, 4 not applicable, 5 registered outside the pinned model.
+`2014-10-31` — 27 of 168 modelled operations implemented (16.1%), 0 stubbed, 4 not applicable, 5 registered outside the pinned model.

 ### Not implemented

@@ -1095,6 +1117,7 @@ signing-certificates: X.509 signing certificates are an EC2-Classic SOAP credent
 - `ApplyPendingMaintenanceAction`
 - `AuthorizeDBSecurityGroupIngress`
 - `BacktrackDBCluster`
+- `CancelDBCapacityReservation`
 - `CancelExportTask`
 - `CopyDBClusterParameterGroup`
 - `CopyDBClusterSnapshot`
@@ -1103,6 +1126,7 @@ signing-certificates: X.509 signing certificates are an EC2-Classic SOAP credent
 - `CopyOptionGroup`
 - `CreateBlueGreenDeployment`
 - `CreateCustomDBEngineVersion`
+- `CreateDBCapacityReservation`
 - `CreateDBClusterEndpoint`
 - `CreateDBClusterParameterGroup`
 - `CreateDBClusterSnapshot`
@@ -1133,6 +1157,7 @@ signing-certificates: X.509 signing certificates are an EC2-Classic SOAP credent
 - `DescribeAccountAttributes`
 - `DescribeBlueGreenDeployments`
 - `DescribeCertificates`
+- `DescribeDBCapacityReservations`
 - `DescribeDBClusterAutomatedBackups`
 - `DescribeDBClusterBacktracks`
 - `DescribeDBClusterEndpoints`
@@ -1174,6 +1199,7 @@ signing-certificates: X.509 signing certificates are an EC2-Classic SOAP credent
 - `ModifyCertificates`
 - `ModifyCurrentDBClusterCapacity`
 - `ModifyCustomDBEngineVersion`
+- `ModifyDBCapacityReservation`
 - `ModifyDBClusterEndpoint`
 - `ModifyDBClusterParameterGroup`
 - `ModifyDBClusterSnapshotAttribute`
@@ -163,11 +163,14 @@ Open between all hosts:
 | Protocol and port | Used by |
 |---|---|
 | UDP 6081 | Geneve tunnels |
+| UDP 500, 4500 and IP protocol 50 (ESP) | IPsec for the Geneve tunnels, while IPsec is on |
 | TCP 4222, 4248 | NATS |
 | TCP 6641, 6642 | OVN northbound and southbound |
 | TCP 8443 | Predastore S3 gate |
 | UDP 6660, 7660 | Object shards, metadata consensus |

+UDP 4500 carries the tunnel traffic only where IKE detects NAT between two nodes; otherwise it travels as ESP, on Oracle Linux as on Debian.
+
 Open between the three OVN database servers (servers 1 to 3) only:

 | Protocol and port | Used by |
@@ -207,7 +210,10 @@ Spinifex's own **host** firewall is separate, and already carries this policy

 Choose one method and apply it to **every** server in the cluster.

-**Option A — existing OS.** On a server already running Ubuntu 26.04 or Debian 13 (Oracle Linux runs a single node only for now, see [Oracle Linux Hosts](/docs/install-oracle-linux)):
+**Option A — existing OS.** Use this on servers already running Ubuntu 26.04, Debian 13 or Oracle Linux 10, and prepare an Oracle Linux server with [Oracle Linux Hosts](/docs/install-oracle-linux) first.
+Every server in a cluster must run the same OS: Oracle Linux never shares a cluster with Debian or Ubuntu, and `spx admin join` does not check this for you.
+
+Run the installer on each server:

 ```bash
 curl -fsSL https://install.mulgadc.com | bash
@@ -1,12 +1,13 @@
 ---
 title: "Oracle Linux Hosts"
 seoTitle: "Install Spinifex on Oracle Linux 10 — Spinifex Docs"
-description: "Prepare an Oracle Linux 10 host for Spinifex: UEK R8, SELinux, a NetworkManager br-wan bridge, the CentOS NFV SIG repository, firewalld handover and updates."
+description: "Prepare Oracle Linux 10 hosts for Spinifex: UEK R8, SELinux, a NetworkManager br-wan bridge, the NFV SIG repository, IPsec on libreswan and host updates."
 category: "Install"
 tags:
  - install
  - oracle linux
  - single node
+  - multi node
 resources:
  - title: "Oracle Linux 10 Documentation"
    url: "https://docs.oracle.com/en/operating-systems/oracle-linux/10/"
@@ -35,13 +36,12 @@ Follow it alongside [Single-Node Install](/docs/install).

 **Supported on Oracle Linux:**

- A single node.
+- A single node, or a cluster in which every node runs Oracle Linux; servers beyond the third join as compute nodes, as on Debian.
 - Guests, VPCs, routed NAT, Elastic IPs, and Application and Network Load Balancers.
 - Whole-GPU passthrough, on x86_64 GPU hosts (see [GPU Passthrough](/docs/gpu-passthrough)).

 **Not supported on Oracle Linux yet:**

- Clusters of more than one node. They wait for IPsec on libreswan, so `spx admin init` turns IPsec off on Oracle Linux and refuses `--ipsec=true`.
 - Clusters that mix Oracle Linux with Debian or Ubuntu nodes, on any release.
 - SELinux enforcing, aarch64, Oracle Linux 9 or earlier, the Red Hat Compatible Kernel, a ZFS root, and NVIDIA MIG.

@@ -51,7 +51,7 @@ Spinifex installs `openvswitch3.5` and `ovn25.03`, the versions Debian 13 ships,
 QEMU comes from Oracle's KVM Utilities repository, because only that build has the `microvm` machine type Spinifex load balancers boot on.

 An Oracle Linux node must be a dedicated Spinifex host.
-Spinifex claims every `tap*` device and stops libreswan's `ipsec.service`, so VPNs, hand-run QEMU guests and libreswan tunnels of your own cannot share the host.
+Spinifex claims every `tap*` device and owns libreswan's `ipsec.service`, which it keeps stopped on a single node and runs for its own tunnels on a cluster node, so VPNs, hand-run QEMU guests and libreswan tunnels of your own cannot share the host.
 Crypto policies other than `DEFAULT` and `FIPS` are untested, and so is Secure Boot.

 ## Prerequisites
@@ -134,8 +134,28 @@ Run the installer as a user with `sudo`, exactly as on Debian:
 curl -fsSL https://install.mulgadc.com | bash
 ```

-Then continue with [Single-Node Install](/docs/install) from Step 2: `setup-ovn.sh --management`, `spx admin init` and `systemctl start spinifex.target`.
-`spx admin init` prints that IPsec is off on this host; on a single node there is no peer to encrypt traffic to.
+For a single node, continue with [Single-Node Install](/docs/install) from Step 2: `setup-ovn.sh --management`, `spx admin init` and `systemctl start spinifex.target`.
+IPsec stays off on a single node, because there is no peer to encrypt traffic to.
+
+For a cluster, prepare and install every node this way, then continue with [Multi-Node Install](/docs/install-multi-node) from Step 2.
+
+### Clusters
+
+Every node in an Oracle Linux cluster runs Oracle Linux.
+`install-node.sh` refuses a host list that mixes Oracle Linux with Debian or Ubuntu, but `spx admin join` does not check, so keep to one OS yourself when joining by hand.
+
+IPsec encrypts the Geneve tunnels between nodes and is on by default, as on Debian.
+On Oracle Linux it runs on libreswan, the IKE daemon Oracle ships, rather than strongSwan.
+Once the node's IPsec certificate exists, the daemon unmasks and starts `ipsec.service` and `openvswitch-ipsec.service`.
+`ovs-monitor-ipsec` then loads the node certificate, its key and the cluster CA into libreswan's NSS database at `/var/lib/ipsec/nss`, under nicknames that start with `ovs_cert`.
+
+The tunnels are plain ESP (IP protocol 50), as on Debian, negotiated over UDP 500.
+Open UDP 500, UDP 4500 and ESP between nodes: libreswan moves to UDP 4500 if it detects NAT between two nodes.
+The Spinifex host firewall opens all three to cluster peers; a firewall between the nodes must allow them too.
+While IKE to a peer fails, the node drops its Geneve traffic to that peer rather than send it in clear.
+The nft table `inet spinifex_ipsec_guard` drops outbound UDP 6081 from the encap address whose route carries no SA, which also covers the moments pluto flushes the xfrm policy table as it starts and stops.
+It is not persisted: `ipsec.service` puts it back at boot before ovn-controller builds the tunnels, and the daemon re-syncs it every 60 seconds.
+`spx admin init --ipsec=false` leaves the tunnels unencrypted, as on any OS.

 ## Step 2. Know What the Installer Changed

@@ -182,7 +202,9 @@ Those advisories do not cover the NFV SIG packages, so follow Open vSwitch and O

 ## Step 5. Uninstall

-`/usr/local/share/spinifex/uninstall-spx.sh` removes Spinifex as on Debian, and lists what it leaves in place on Oracle Linux: the repository changes and the SIG key, `OVS_USER_ID` and `OVN_USER_ID`, the OVS and OVN logrotate edits, the chrony include, `/etc/modules-load.d/spinifex.conf`, the SELinux mode, a masked `ipsec.service`, firewalld disabled, and any GPU passthrough configuration.
+`/usr/local/share/spinifex/uninstall-spx.sh` removes Spinifex as on Debian.
+On a cluster node it also stops `openvswitch-ipsec.service` and deletes the `ovs_cert*` entries from libreswan's NSS database, so the node can join a different cluster; other NSS identities and your connection files in `/etc/ipsec.d` stay.
+It lists what it leaves in place on Oracle Linux: the repository changes and the SIG key, `OVS_USER_ID` and `OVN_USER_ID`, the OVS and OVN logrotate edits, the chrony include, `/etc/modules-load.d/spinifex.conf`, the SELinux mode, a masked `ipsec.service`, firewalld disabled, and any GPU passthrough configuration.
 `--purge-deps` is refused on Oracle Linux, because dnf cannot tell the packages Spinifex installed from the host's own; remove them with `dnf remove` yourself.

 ## Troubleshooting
@@ -224,6 +246,20 @@ If it prints `yes`, run `sudo nmcli connection modify br-wan bridge.stp no` and

 Check that `/usr/bin/spx` links to `/usr/local/bin/spx`; run the installer again if it does not, or call `sudo /usr/local/bin/spx` meanwhile.

+### Cluster nodes form no IPsec tunnels
+
+Check that both IPsec services run, and that libreswan holds the node's certificate and the cluster CA:
+
+```bash
+systemctl is-active ipsec.service openvswitch-ipsec.service
+sudo certutil -L -d sql:/var/lib/ipsec/nss     # ovs_certkey_<node>, ovs_cert_cacert
+sudo ipsec trafficstatus                       # one line per peer tunnel
+```
+
+If the services are inactive, check the daemon log with `journalctl -u spinifex-daemon` for the IPsec reconcile.
+If the certificates are present but no tunnel is listed, check that UDP 500 and ESP pass between the nodes in both directions, and look for `SEC_ERROR_UNKNOWN_ISSUER` in `journalctl -u ipsec`, which means a peer's certificate is from another CA.
+Until the tunnel forms, guest traffic between those two nodes is dropped, which is the guard in [Clusters](#clusters) working.
+
 ### CA certificate not trusted

 `sudo spx admin init` installs the CA automatically. To install it again by hand on Oracle Linux:
@@ -39,7 +39,7 @@ This guide installs Spinifex on a single server using the binary installer. For

 - Ubuntu 26.04
 - Debian 13
- Oracle Linux 10.2 or a later 10.x, on UEK R8, x86_64 only, as a single node. Open vSwitch and OVN come from the CentOS NFV SIG, a community repository that Oracle does not support. Prepare the host with [Oracle Linux Hosts](/docs/install-oracle-linux) first.
+- Oracle Linux 10.2 or a later 10.x, on UEK R8, x86_64 only, as a single node or, with [Multi-Node Install](/docs/install-multi-node), a cluster. Open vSwitch and OVN come from the CentOS NFV SIG, a community repository that Oracle does not support. Prepare the host with [Oracle Linux Hosts](/docs/install-oracle-linux) first.

 **What Gets Installed:**

@@ -99,9 +99,6 @@ behavior in other words.
 | 4432 | Formation server | HTTPS | Cluster (bootstrap only) | Cluster join coordination; active only while a join token is valid. Binds the node's `--bind` (lan) address. See *Formation port lifecycle* below. | Short-lived bearer token + TLS¹ |
 | 4222 | spinifex-nats (client) | NATS + TLS | Cluster | Internal service bus for EC2/EBS/VPC/S3 handlers | Token + mutual TLS (cluster CA) |
 | 4248 | spinifex-nats (cluster) | NATS + TLS | Cluster | Inter-node NATS federation | Token + mutual TLS (cluster CA) |
-| 4223 | spinifex-sqs-store (client) | NATS + TLS | Cluster | SQS message data, separate from the control-plane bus | Store-only token + TLS |
-| 4249 | spinifex-sqs-store (cluster) | NATS + TLS | Cluster | Inter-node SQS store replication | Store-only token + mutual TLS (cluster CA) |
-| 8223 | spinifex-sqs-store (monitor) | HTTP | Loopback | varz and jsz for the store | Loopback only |
 | 5300 | northstar | DNS (UDP + TCP) | Cluster | Forward target for every node's per-instance DNS shim, dialled cross-node. Binds the wildcard by design. | None — restrict by firewall |
 | 6660 | predastore (blob node) | QUIC / UDP | Cluster | Erasure-coded object shard transport between hosts. Multi-node clusters only — see *Predastore ports* below. | TLS 1.3, server certificate verified against the cluster CA |
 | 7660 | predastore (meta node) | QUIC / UDP | Cluster | Raft consensus over global state — buckets and the object index — between hosts. Multi-node clusters only. | TLS 1.3, server certificate verified against the cluster CA |
@@ -111,8 +108,8 @@ behavior in other words.
 | 6643 | OVN Northbound DB (RAFT) | OVSDB/TCP | Cluster | NB database RAFT replication between the 3 quorum nodes | Cluster network only; TLS planned |
 | 6644 | OVN Southbound DB (RAFT) | OVSDB/TCP | Cluster | SB database RAFT replication between the 3 quorum nodes | Cluster network only; TLS planned |
 | 6081 | OVN (Geneve) | UDP | Encap | Tenant traffic overlay between chassis. A kernel UDP-tunnel socket, so packets are delivered locally and traverse the host's netfilter input hook before OVS sees them — a host firewall must accept them explicitly. The socket is opened by the kernel tunnel driver and takes no bind address, so it binds the wildcard by design and reach must be restricted by firewall. | None — see 500/4500 |
-| 500, 4500 | strongSwan `charon` | IKEv2 / UDP | Encap | IKE and NAT-T for the IPsec protecting Geneve, managed entirely by `ovs-monitor-ipsec`. Binds the wildcard by design (the upstream strongSwan default, accepted rather than overridden), so reach must be restricted by firewall. | Certificate-based, against the cluster CA |
-| — | ESP | IP proto 50 | Encap | The IPsec payload itself, once IKE has negotiated an SA | Cluster CA |
+| 500, 4500 | strongSwan `charon`, or libreswan `pluto` on Oracle Linux | IKEv2 / UDP | Encap | IKE and NAT-T for the IPsec protecting Geneve, managed entirely by `ovs-monitor-ipsec`; 4500 carries the ESP payload only where IKE detects NAT between two nodes. `charon` binds the wildcard by design (the upstream strongSwan default, accepted rather than overridden); `pluto` binds each host address separately, the WAN address included, which is no narrower. Either way reach must be restricted by firewall. | Certificate-based, against the cluster CA |
+| — | ESP | IP proto 50 | Encap | The IPsec payload itself, once IKE has negotiated an SA, on Oracle Linux as on Debian. | Cluster CA |
 | 8222 | spinifex-nats (monitoring) | HTTP | Localhost | `varz`/`subsz` metrics consumed by the daemon | Loopback only |
 | 323 | chronyd | NTP | Localhost | Time sync client control socket | Loopback only |
 | 169.254.169.254:80 | spinifex-vpcd (IMDS) | HTTP | Guest | Instance metadata service. One socket per instance, bound to that instance's `ime-*` interface. Terminates on the host, so it traverses the netfilter input hook. | Instance identity by interface; IMDSv2 tokens |
@@ -177,7 +174,7 @@ Control-plane and data-plane traffic between Spinifex nodes, for completeness an
 | OVN NB/SB (client) | 6641 / 6642 | Cluster network only (TLS planned) | Network control plane; vpcd and ovn-controller dial the quorum |
 | OVN NB/SB (RAFT) | 6643 / 6644 | Cluster network only (TLS planned) | NB/SB database replication across the 3 quorum nodes |
 | OVN tunnels (Geneve) | UDP 6081 | Encapsulated by IPsec when `network.ipsec_enabled` is true, which is the default on multi-node clusters | Tenant traffic overlay between chassis, on the `vpc` plane |
-| OVN IPsec (IKE / NAT-T / ESP) | UDP 500, UDP 4500, IP proto 50 | Certificate-based against the cluster CA, negotiated by `ovs-monitor-ipsec` | Protects the Geneve tunnels above. OVN-native only — no layer manages strongSwan directly. |
+| OVN IPsec (IKE / NAT-T / ESP) | UDP 500, UDP 4500, IP proto 50 | Certificate-based against the cluster CA, negotiated by `ovs-monitor-ipsec` | Protects the Geneve tunnels above. OVN-native only: no layer configures strongSwan or libreswan directly. On Oracle Linux the daemon also blocks outbound Geneve that no SA covers, so a peer failing IKE gets no plaintext. |
 | Instance DNS forwarding | UDP/TCP 5300 | None | Each node's per-instance DNS shim forwards guest queries to peer nodes' northstar `:5300` |

 Nodes **must** sit on a network segment that is not routed to tenant/guest VMs or to the internet. The Predastore blob and meta transports and the OVN DBs are cluster-internal and must not be reachable from anywhere else.
@@ -224,7 +221,7 @@ table inet spinifex_filter {

    # Cluster, from peer nodes only. Replace with your nodes' lan-plane
    # addresses — a CIDR does not generalise to nodes on different subnets.
-    ip saddr { 10.0.1.1, 10.0.1.2, 10.0.1.3 } tcp dport { 4222, 4223, 4248, 4249, 4432, 5300, 6641, 6642, 6643, 6644, 8660 } accept
+    ip saddr { 10.0.1.1, 10.0.1.2, 10.0.1.3 } tcp dport { 4222, 4248, 4432, 5300, 6641, 6642, 6643, 6644, 8660 } accept
    ip saddr { 10.0.1.1, 10.0.1.2, 10.0.1.3 } udp dport { 5300, 6660, 7660 } accept

    # Encap, from peer chassis only. Replace with your nodes' vpc-plane
@@ -251,7 +248,7 @@ Every listener and outbound destination is controlled by one of these files. Cha

 | File | Keys | Controls |
 |------|------|----------|
-| `/etc/spinifex/spinifex.toml` | `nodes.<node>.{awsgw,nats,predastore,daemon}.host`, `nodes.<node>.vpcd.ovn_{nb,sb}_addr`, `nodes.<node>.daemon.dev_networking`, `network.ipsec_enabled` | Per-service bind addresses/ports; dev-mode QEMU port forwarding. `network.ipsec_enabled` (default `true`) decides whether OVN-native IPsec protects the Geneve tunnels, and therefore whether `charon` listens on 500/4500 — single-node clusters never enable it. |
+| `/etc/spinifex/spinifex.toml` | `nodes.<node>.{awsgw,nats,predastore,daemon}.host`, `nodes.<node>.vpcd.ovn_{nb,sb}_addr`, `nodes.<node>.daemon.dev_networking`, `network.ipsec_enabled` | Per-service bind addresses/ports; dev-mode QEMU port forwarding. `network.ipsec_enabled` (default `true`) decides whether OVN-native IPsec protects the Geneve tunnels, and therefore whether the IKE daemon (`charon`, or `pluto` on Oracle Linux) listens on 500/4500 — single-node clusters never enable it. |
 | `/etc/spinifex/nats.conf` | `listen`, `cluster.listen`, `cluster.routes`, `http`, `tls`, `cluster.authorization` | NATS client/cluster/monitoring listeners, peer routes, TLS, cluster token. |
 | `/etc/spinifex/predastore/predastore.toml` | `[[host]].bind_addr`, `[[host]].addr`, `[[host]].tls_cert`, `[[host]].tls_key`, `[[host.node]].role`, `[[host.node]].port` | Predastore host and node layout: `bind_addr` is the address the host's sockets bind, `addr` is the address peer hosts dial it on, and both carry no port — the nodes pinned to the host supply their own. The service's `--host`/`--port` override the bind address and the gate's S3 port. |
 | `/etc/spinifex/northstar/northstar.toml` | `listen`, `forwarders` | DNS listen addresses (`:53` on the advertise address, `:5300` wildcard) and upstream resolvers. |
@@ -1,521 +0,0 @@
-# Service interface manifest.
-#
-# Source of truth for "which services produce/consume which NATS subjects"
-# and "which e2e suites cover which services". Consumed by
-# tests/e2e/select-suites (Bead 4) to derive targeted suite selection from
-# a PR diff.
-#
-# See docs/development/improvements/e2e-targeted-suite-selection.md.
-#
-# Editing rules
-# -------------
-# - subscribes / publishes: subjects this service receives / sends. Both
-#   lists use the literal subject string, or a "<prefix>.*" wildcard for
-#   pattern subscriptions (e.g. "ebs.snapshot.*").
-# - depends_on: subjects this service requires from peers. Used by the
-#   selector to compute "if X changes, which suites might break".
-# - Subjects must match what is actually referenced in source. The lint
-#   in Bead 5 greps `"(ec2|vpc|ebs|s3|iam|elbv2|spinifex)\.[a-zA-Z._-]+"`
-#   and fails if any subject is absent from this manifest.
-
-version: 1
-
-services:
-
-  awsgw:
-    path: spinifex/runtime/roles/awsgw
-    # TLS frontend, SigV4 auth. Pure router: HTTP → NATS request/response.
-    # No subscribes; only publishes (via NATSRequest).
-    subscribes: []
-    publishes:
-      - "ec2.*"
-      - "elbv2.*"
-      - "acm.*"
-      - "route53.*"
-      - "iam.*"
-      - "rds.*"
-      - "ecs.*"
-      - "ochre.vector.*"
-    depends_on:
-      - "ec2.*"
-      - "elbv2.*"
-      - "acm.*"
-      - "route53.*"
-      - "iam.*"
-      - "rds.*"
-      - "ecs.*"
-      - "ochre.vector.*"
-
-  spinifex_daemon:
-    path: spinifex/runtime/roles/spinifex
-    # Orchestrator. Owns most ec2.* + all elbv2.* + iam event publication.
-    # Translates VPC / EBS / S3 operations into vpc.* / ebs.* / s3.* and
-    # forwards to vpcd / viperblockd / predastore.
-    #
-    # Source layout: spinifex/runtime/roles/spinifex (entry), spinifex/runtime/daemon
-    # (subscribe registration), spinifex/domains/<domain> and the legacy
-    # spinifex/handlers/{ecs,eks,elbv2,iam,rds} packages (per-subject handler
-    # bodies). Selector treats these paths as belonging to this service via
-    # the `additional_paths` field.
-    additional_paths:
-      - spinifex/runtime/daemon
-      - spinifex/domains/ec2/account
-      - spinifex/domains/ec2/eigw
-      - spinifex/domains/ec2/eip
-      - spinifex/domains/ec2/igw
-      - spinifex/domains/ec2/image
-      - spinifex/domains/ec2/instance
-      - spinifex/domains/ec2/key
-      - spinifex/domains/ec2/launchtemplate
-      - spinifex/domains/ec2/natgw
-      - spinifex/domains/ec2/placementgroup
-      - spinifex/domains/ec2/routetable
-      - spinifex/domains/ec2/snapshot
-      - spinifex/domains/ec2/spotinstance
-      - spinifex/domains/ec2/tags
-      - spinifex/domains/ec2/volume
-      - spinifex/domains/ec2/volumestate
-      - spinifex/domains/ec2/vpc
-      - spinifex/domains/elbv2
-      - spinifex/domains/acm
-      - spinifex/domains/route53
-      - spinifex/domains/iam
-      - spinifex/domains/ecs
-      - spinifex/domains/rds/instance
-      - spinifex/domains/rds/subnetgroup
-      - spinifex/domains/rds/parametergroup
-      - spinifex/domains/rds/engine
-      - spinifex/domains/ecs/taskdefinition
-      - spinifex/domains/eks/cluster
-      - spinifex/domains/eks/access
-      - spinifex/domains/eks/addon
-      # Guest add-on wire contract (subject, report, staged-manifest body).
-      - contracts/eks/v1
-      # ECS agent/scheduler Layer-2 bus wire contract (subjects, message types).
-      - contracts/ecs/v1
-      - spinifex/domains/ochre/vector
-    subscribes:
-      # EC2 instance lifecycle.
-      - "ec2.RunInstances"
-      - "ec2.DescribeInstances"
-      - "ec2.DescribeStoppedInstances"
-      - "ec2.DescribeTerminatedInstances"
-      - "ec2.DescribeInstanceAttribute"
-      - "ec2.DescribeInstanceStatus"
-      - "ec2.DescribeInstanceTypes"
-      - "ec2.ModifyInstanceAttribute"
-      - "ec2.ModifyVolume"
-      - "ec2.start"
-      - "ec2.terminate"
-      - "ec2.launch"
-      # contracts/ec2/v1 owns the command payload and this subject vocabulary.
-      # The wildcard is intentionally one final instance-ID token, not `>`.
-      - "ec2.cmd.*"
-      # EC2 VPC / network ops handled by daemon before delegation to vpcd.
-      - "ec2.CreateVpc"
-      - "ec2.DescribeVpcs"
-      - "ec2.CreateSubnet"
-      - "ec2.CreateInternetGateway"
-      - "ec2.CreateKeyPair"
-      - "ec2.CreateImage"
-      - "ec2.DescribePlacementGroups"
-      # Cluster coordination subjects.
-      - "ec2.ReserveClusterNode"
-      - "ec2.FinalizeClusterInstances"
-      # An EIP association reaches whichever node holds the instance, which is
-      # rarely the node that served the request (fan-out, no queue group).
-      - "ec2.eni.public-ip-changed"
-      # contracts/cluster/v1 owns the node status/capacity payload and subject.
-      - "spinifex.node.status"
-      # contracts/cluster/v1 owns the VM inventory payload and subject.
-      - "spinifex.node.vms"
-      - "spinifex.nodes.discover"
-      - "spinifex.cluster.shutdown.progress"
-      # Coordinated cluster-shutdown phases (fan-out, no queue group).
-      - "spinifex.cluster.shutdown.gate"
-      - "spinifex.cluster.shutdown.drain"
-      - "spinifex.cluster.shutdown.storage"
-      - "spinifex.cluster.shutdown.persist"
-      - "spinifex.cluster.shutdown.infra"
-      # Operator-plane Predastore topology contract (contracts/operator/v1).
-      - "spinifex.storage.config"
-      # ELBv2 control plane (domains/elbv2/service_nats.go topic map).
-      - "elbv2.CreateLoadBalancer"
-      - "elbv2.DeleteLoadBalancer"
-      - "elbv2.DescribeLoadBalancers"
-      - "elbv2.CreateTargetGroup"
-      - "elbv2.DeleteTargetGroup"
-      - "elbv2.DescribeTargetGroups"
-      - "elbv2.RegisterTargets"
-      - "elbv2.DeregisterTargets"
-      - "elbv2.DescribeTargetHealth"
-      - "elbv2.CreateListener"
-      - "elbv2.DeleteListener"
-      - "elbv2.DescribeListeners"
-      # ACM-alike certificate store (domains/acm; subscribed in daemon.go).
-      - "acm.ImportCertificate"
-      - "acm.DescribeCertificate"
-      - "acm.ListCertificates"
-      - "acm.DeleteCertificate"
-      # Route 53 hosted zones (domains/route53; subscribed in daemon.go).
-      - "route53.*"
-      # Operator-only zone activation and delegation, sent by spx admin dns.
-      - "route53.admin.*"
-      - "elbv2.AddListenerCertificates"
-      - "elbv2.RemoveListenerCertificates"
-      - "elbv2.DescribeListenerCertificates"
-      - "elbv2.DescribeSSLPolicies"
-      # RDS control plane (domains/rds/instance/subjects.go). One wildcard rather than
-      # 30 literals: every action subject is registered from a constant, so a
-      # per-subject list would be a hand-mirror of that file with nothing
-      # checking it.
-      - "rds.*"
-      # ECS control plane (domains/ecs/service_nats.go, registered in
-      # daemon.go). One wildcard, as for RDS: every action subject is a literal
-      # in those two files, so a per-subject list would only mirror them.
-      - "ecs.*"
-      # Layer-2 agent bus: register and health are wildcard subscriptions over
-      # {account}.{db-instance}, so the subject is built at runtime and only the
-      # prefix is declarable.
-      - "rds.bus.*"
-      # ECS Layer-2 scheduler/agent bus (contracts/ecs/v1): account, cluster and
-      # instance/task IDs are runtime-built, so only the prefix is declarable.
-      # Only the leader holds these subscriptions, so there is no queue group.
-      - "ecs.bus.*"
-      # Ochre vector store tenant surface (domains/ochre/vector/api.go),
-      # gated behind config.OchreVectorConfig.Enabled (default off).
-      - "ochre.vector.createIndex"
-      - "ochre.vector.deleteIndex"
-      - "ochre.vector.listIndexes"
-      - "ochre.vector.ingest"
-      - "ochre.vector.describeJob"
-      - "ochre.vector.query"
-      - "ochre.vector.listJobs"
-      - "ochre.vector.stopJob"
-      # Fence route owned by contracts/viperblockd/legacy/v1: viperblockd
-      # announces a torn-down export so this node stops the guest that was
-      # using it. "ebs.fenced" is the single-node form; a named node is
-      # addressed as "ebs.<node>.fenced".
-      - "ebs.fenced"
-    publishes:
-      # → vpcd
-      - "vpc.create"
-      - "vpc.delete"
-      - "vpc.create-subnet"
-      - "vpc.delete-subnet"
-      - "vpc.create-sg"
-      - "vpc.delete-sg"
-      - "vpc.update-sg"
-      - "vpc.create-port"
-      - "vpc.delete-port"
-      - "vpc.update-port-sgs"
-      # EC2-to-network projection owned by contracts/ec2/v1. The network
-      # subscriber realizes it in OVN without reading EC2 private state.
-      - "vpc.igw-attach"
-      - "vpc.igw-detach"
-      - "vpc.add-nat-gateway"
-      - "vpc.delete-nat-gateway"
-      - "vpc.add-nat"
-      - "vpc.delete-nat"
-      - "vpc.dhcp.acquire"
-      - "vpc.dhcp.release"
-      # Per-subnet route-table projections owned by domains/ec2/routetable:
-      # IGW default-route reachability and the egress gate/ungate decision it
-      # implies, published fire-and-forget on every route-table mutation.
-      - "vpc.add-igw-route"
-      - "vpc.delete-igw-route"
-      - "vpc.gate-subnet-egress"
-      - "vpc.ungate-subnet-egress"
-      # Egress-only reroute removal for a single system instance
-      # (domains/eks/cluster). Fire-and-forget. vpc.add-system-egress has no
-      # current producer.
-      - "vpc.delete-system-egress"
-      # → viperblockd
-      # The migrated legacy ebs.config and ebs.sync payloads and routes are
-      # owned by contracts/viperblockd/legacy/v1. They are transitional and
-      # distinct from the generic providers/ebs protocol; other legacy ebs.*
-      # routes remain to be migrated separately.
-      - "ebs.mount"
-      - "ebs.unmount"
-      - "ebs.delete"
-      - "ebs.sync"
-      - "ebs.snapshot.*"
-      - "ebs.config"
-      - "ebs.config.*"
-      # Abandon route owned by contracts/viperblockd/legacy/v1: the daemon
-      # asks viperblockd to tear an export down without sealing. "ebs.abandon"
-      # is the single-node form; a named node addresses "ebs.<node>.abandon".
-      - "ebs.abandon"
-      # → predastore (S3 / AMI / keypair store)
-      - "s3.*"
-      # → every daemon, so the one holding the instance reports the new address
-      - "ec2.eni.public-ip-changed"
-      # IAM lifecycle event
-      - "iam.account.created"
-      # → the in-guest rds-agent: set-password and the like, published on the
-      # per-instance command subject and answered on its reply subject.
-      - "rds.bus.*"
-      # ECS service-reconcile self-wake (contracts/ecs/v1): the worker that
-      # recorded a task-state change is rarely the leader, so this fans the
-      # wake out and the leader's own "ecs.bus.*" subscription picks it up.
-      - "ecs.bus.*"
-    depends_on:
-      - "vpc.*"
-      - "ebs.*"
-      - "s3.*"
-
-  vpcd:
-    path: spinifex/runtime/roles/vpcd
-    # OVN-backed VPC / subnet / SG / NAT / IGW data plane.
-    subscribes:
-      - "vpc.create"
-      - "vpc.delete"
-      - "vpc.create-subnet"
-      - "vpc.delete-subnet"
-      - "vpc.create-sg"
-      - "vpc.delete-sg"
-      - "vpc.update-sg"
-      - "vpc.create-port"
-      - "vpc.delete-port"
-      - "vpc.update-port-sgs"
-      - "vpc.igw-attach"
-      - "vpc.igw-detach"
-      - "vpc.add-nat-gateway"
-      - "vpc.delete-nat-gateway"
-      - "vpc.add-nat"
-      - "vpc.delete-nat"
-      - "vpc.dhcp.acquire"
-      - "vpc.dhcp.release"
-      - "vpc.add-igw-route"
-      - "vpc.delete-igw-route"
-      - "vpc.gate-subnet-egress"
-      - "vpc.ungate-subnet-egress"
-      - "vpc.add-system-egress"
-      - "vpc.delete-system-egress"
-    publishes:
-      - "vpc.port-status"
-      - "vpc.dhcp.lease-changed"
-      - "vpc.dhcp.owner-check"
-    depends_on: []
-
-  viperblockd:
-    path: spinifex/runtime/roles/viperblockd
-    # EBS data plane. Backed by the viperblock storage engine in the
-    # mulga/viperblock submodule. Submodule pointer bump → selector forces
-    # "run all" (contract changes can be invisible at the spinifex source
-    # level).
-    # The migrated legacy ebs.config and ebs.sync routes are versioned by
-    # contracts/viperblockd/legacy/v1, not by providers/ebs. Other legacy
-    # ebs.* routes remain to be migrated separately.
-    # The dirty-marker JetStream KV bucket VIPERBLOCK_VOLUME_DIRTY is owned
-    # by the same contract; not a NATS subject, so not listed below.
-    #
-    # Source layout: spinifex/runtime/roles/viperblockd (role: process lifecycle,
-    # NATS subscription wiring) and spinifex/providers/ebs/viperblock (the
-    # Viperblock EBS provider adapter the role wires up). Selector treats the
-    # adapter path as belonging to this service via `additional_paths`.
-    additional_paths:
-      - spinifex/providers/ebs/viperblock
-    subscribes:
-      - "ebs.mount"
-      - "ebs.unmount"
-      - "ebs.delete"
-      - "ebs.sync"
-      - "ebs.snapshot.*"
-      - "ebs.config"
-      - "ebs.config.*"
-      # Abandon route (contracts/viperblockd/legacy/v1): request/reply from
-      # the daemon, answered by the Viperblock adapter on the named node.
-      - "ebs.abandon"
-    publishes:
-      - "ebs.mount.response"
-      # Fence route (contracts/viperblockd/legacy/v1): announced after the
-      # Viperblock adapter tears an export down because its lease moved.
-      - "ebs.fenced"
-      - "ebs.unmount.response"
-    depends_on: []
-
-  predastore:
-    path: spinifex/runtime/roles/predastore
-    # S3 / AMI / keypair storage. External submodule under mulga/predastore;
-    # this dir contains only the spinifex-side adapter. Submodule pointer
-    # bump → selector forces "run all".
-    subscribes:
-      - "s3.*"
-    publishes: []
-    depends_on: []
-
-  nats:
-    path: spinifex/runtime/roles/nats
-    # NATS broker bootstrap / cluster config. Not a NATS pub/sub
-    # participant. Any change here is infra-path; selector forces
-    # "run all".
-    subscribes: []
-    publishes: []
-    depends_on: []
-
-  spinifexui:
-    path: spinifex/runtime/roles/spinifexui
-    # Web UI. Talks to awsgw over HTTP, never NATS. Diffs limited to this
-    # dir do not need any e2e suite to run.
-    subscribes: []
-    publishes: []
-    depends_on: []
-    covers_no_suite: true
-
-
-# Fixture → service mapping. Each Ensure* helper in
-# tests/e2e/harness/fixtures.go exercises this set of services; the
-# Bead 4 selector AST-scans test files for harness.Ensure* call sites
-# and resolves coverage via this block.
-fixtures:
-
-  EnsureKeyPair:
-    services: [spinifex_daemon, predastore]
-
-  EnsureAMI:
-    services: [spinifex_daemon, predastore, viperblockd]
-
-  EnsureDefaultVPC:
-    services: [vpcd, spinifex_daemon]
-
-  EnsureSubnet:
-    services: [vpcd, spinifex_daemon]
-
-  EnsureSG:
-    services: [vpcd, spinifex_daemon]
-
-  EnsureInstance:
-    services: [spinifex_daemon, viperblockd, vpcd]
-
-  EnsureVolume:
-    services: [spinifex_daemon, viperblockd]
-
-  EnsureSnapshot:
-    services: [spinifex_daemon, viperblockd, predastore]
-
-  EnsureNATGateway:
-    services: [vpcd, spinifex_daemon]
-
-  EnsureLoadBalancer:
-    services: [spinifex_daemon, awsgw]
-
-
-suites:
-
-  e2e-cert:
-    path: tests/e2e/cert
-    covers: [awsgw, spinifex_daemon, vpcd, predastore]
-
-  e2e-lb:
-    path: tests/e2e/lb
-    covers: [awsgw, spinifex_daemon, vpcd, viperblockd, predastore]
-    # Exercises full ELBv2 control plane (domains/elbv2) + data plane
-    # (LB agent on the system VM). ELBv2 is part of spinifex_daemon in
-    # this manifest; revisit if it splits into a standalone service.
-
-  e2e-single:
-    path: tests/e2e/single
-    covers: [awsgw, spinifex_daemon, vpcd, viperblockd, predastore]
-    # Full single-node lifecycle: VPC, AMI, keypair, instance, volume,
-    # snapshot, IAM. Currently runs all phases sequentially under one
-    # `TestSingleNode`. Bead 3 (single-fixture-migration) splits this
-    # into per-test fixture-driven coverage so the selector can pick
-    # subtests, not just whole suites.
-
-  e2e-iam:
-    path: tests/e2e/iam
-    covers: [awsgw, spinifex_daemon, vpcd, viperblockd, predastore]
-    # "Everything IAM": the IAM/STS control plane plus the instance-identity
-    # tests (instance-profile association + IMDS), which boot guest VMs — OVN
-    # via vpcd, AMI via predastore, root volume via viperblockd, lifecycle via
-    # spinifex_daemon. Single-mode, VM-booting leg.
-
-  e2e-ecr:
-    path: tests/e2e/ecr
-    covers: [awsgw, spinifex_daemon, predastore]
-    # ECR control plane (domains/ecr, gateway/ecr*) + OCI data plane: blobs and
-    # manifests in predastore, control + auth bridge in awsgw, KV via
-    # spinifex_daemon. docker/crane/skopeo data-plane subtests skip when the
-    # client binary or registry-host resolution is absent.
-
-  e2e-bedrock:
-    path: tests/e2e/bedrock
-    covers: [awsgw]
-    # Ochre (AWS-Bedrock) control plane: static catalog list/get served
-    # straight from awsgw (gateway/bedrock.go, domains/ochre/) — no NATS
-    # hop to spinifex_daemon. Converse (Anthropic-direct / self-host vLLM)
-    # subtests are gated on backend availability and skip by default.
-
-  e2e-ecs:
-    path: tests/e2e/ecs
-    covers: [awsgw, spinifex_daemon, vpcd, viperblockd, predastore]
-    # ECS data plane end-to-end: cluster + a real container instance launched
-    # from the spinifex-ecs-node AMI (boots, registers over the gateway, runs
-    # tasks through containerd), standalone tasks in awsvpc/bridge/host modes,
-    # and an awsvpc service fronted by an ELBv2 target group. Boots a guest VM,
-    # so vpcd (OVN + awsvpc ENI), viperblockd (root volume), predastore (AMI)
-    # are all in the path.
-
-  e2e-rds:
-    path: tests/e2e/rds
-    covers: [awsgw, spinifex_daemon, vpcd, viperblockd, predastore]
-    # RDS create/describe/connect: CreateDBInstance boots a dual-NIC DB VM from
-    # the engine AMI (predastore) with a Viperblock data volume (viperblockd) and
-    # a customer ENI (vpcd), and the reconciler flips it to available on the
-    # in-guest agent's first healthy heartbeat over awsgw. Every test deletes the
-    # instances it creates, so the suite needs no gate beyond SPINIFEX_E2E.
-
-  e2e-ddil:
-    path: tests/e2e/ddil
-    covers: [awsgw, spinifex_daemon, vpcd, viperblockd, predastore]
-    # DDIL = degraded / disconnected / intermittent / low-bandwidth.
-    # Exercises the full daemon stack under network failure injection;
-    # any service touching the wire path is in scope.
-
-  e2e-route53:
-    path: tests/e2e/route53
-    covers: [awsgw, spinifex_daemon, predastore]
-    # Hosted zones end to end: delegation check, publish to the northstar zone
-    # store in predastore, and every node answering before a change is INSYNC.
-
-  # Suites below are bash-only today. Entries activate as the Go ports
-  # land (Bead 2 fixtures unblock new-shape ports for multinode etc.).
-  #
-  # e2e-multinode:
-  #   path: spinifex/tests/e2e/multinode
-  #   covers: [awsgw, spinifex_daemon, vpcd, viperblockd, predastore]
-  # e2e-reboot:
-  #   path: spinifex/tests/e2e/reboot
-  #   covers: [spinifex_daemon, viperblockd]
-  # e2e-tofu:
-  #   path: spinifex/tests/e2e/tofu
-  #   covers: [awsgw, spinifex_daemon, vpcd]
-  # e2e-bm:
-  #   path: spinifex/tests/e2e/baremetal
-  #   covers: [awsgw, spinifex_daemon, vpcd, viperblockd, predastore]
-  #
-  # e2e-natuplink activates once CI can provision an external_mode=nat
-  # node (setup-ovn.sh --nat-uplink + spx admin init --external-mode=nat).
-  # e2e-natuplink:
-  #   path: spinifex/tests/e2e/natuplink
-  #   covers: [awsgw, spinifex_daemon, vpcd, viperblockd, predastore]
-
-
-# Special-case rules consumed by the selector (Bead 4). Hard-coded in
-# the tool, listed here for visibility.
-#
-# infra_paths_force_all:
-#   - spinifex/runtime/roles/nats/**
-#   - spinifex/tests/e2e/harness/**
-#   - spinifex/tests/e2e/lib/**
-#   - spinifex/.github/workflows/e2e-go.yml
-#   - spinifex/.github/scripts/**
-#   - spinifex/Makefile
-#   - mulga/scripts/tofu-cluster/**
-#
-# submodule_force_all:
-#   - mulga/predastore (submodule pointer)
-#   - mulga/viperblock (submodule pointer)
-#
-# many_services_force_all_threshold: 3   # ≥ N services touched → run all
@@ -40,7 +40,7 @@ Runs **on** a node, shelling out to `ip`/`iptables`/`ovn-nbctl` locally. Skips u

 Phases 7 and 8 are gated on a cluster by what the datapath can actually do: the transit segment is a localnet, so host-to-guest Tier 1 ingress only exists where the VPC's gateway router port and the guest are both on this host. The VPC ingress route itself is installed on every node, so phase 8 checks it unconditionally.

-Three nightly cells run it: `nat-single` (cell 19), its Oracle Linux twin (cell 36) and `nat-multi` (cell 30). Cell 20 is the bare-metal job and is not in the permutation matrix at all, which is why the multi-node NAT cell is numbered past the end rather than filling a gap.
+Four nightly cells run it: `nat-single` (cell 19) and `nat-multi` (cell 30), and their Oracle Linux twins (cells 36 and 38). Cell 20 is the bare-metal job and is not in the permutation matrix at all, which is why the multi-node NAT cell is numbered past the end rather than filling a gap.

 ### `multinode` — behaviour that only exists on more than one node

@@ -50,6 +50,9 @@ Three nightly cells run it: `nat-single` (cell 19), its Oracle Linux twin (cell

 **`TestMultinodeDNS` convergence budgets are env-tunable.** `SPINIFEX_MULTINODE_DNS_TOPOLOGY_TIMEOUT` (default `90s`) bounds every node agreeing on the NS topology, `SPINIFEX_MULTINODE_DNS_TIMEOUT` (default `2m`) bounds a record becoming answerable from a guest or a node, and `SPINIFEX_MULTINODE_DNS_POLL` (default `3s`) is the poll interval for both. Values are Go durations; a malformed or non-positive value panics rather than falling back to the default.

+**`TestMultinodeOVNRaft` restarts each node's OVN central in turn before it kills the NB leader.**
+A `setup-ovn.sh` rerun restarts a node's NB and SB members together, on EL through the one `ovn-northd.service`, so the test does the same on every database node in turn, waits for both members to be back as cluster members under a known leader, and requires an NB write to land after each step and the full quorum with one leader at the end.
+
 **`TestMultinodeIPSec` proves guest traffic is encrypted on the wire, not only that IPsec is configured.**
 It first requires every node to carry the OVS certificate, key and CA pointers, `NB_Global.ipsec=true` never asserted over an unconfigured chassis, and an AES-GCM SA in `ip xfrm state`.
 None of those fails when SAs exist but traffic bypasses them, so it then launches two instances on distinct nodes and sends 200 pings of 1000 bytes between them.
@@ -140,7 +143,6 @@ The create/describe/connect path against a live cluster: a DB VM that actually b
 - **`ddil`** — Denied, Disrupted, Intermittent, Limited network scenarios.
 - **`teardown`** — sweeps leftovers.
 - **`harness`** — shared primitives, not a suite. `harness/vpc_egress.go` carries the VPC/subnet/IGW/NAT-gateway fixtures the egress scenarios build on.
- **`manifest-check`, `manifest-lint`** — manifest validation, not cluster behaviour.

 ## Keeping this honest

@@ -33,6 +33,7 @@ require (
 	github.com/mulgadc/viperblock v1.23.0
 	github.com/nats-io/nats-server/v2 v2.15.1
 	github.com/nats-io/nats.go v1.54.0
+	github.com/nats-io/nkeys v0.4.16
 	github.com/opencontainers/runtime-spec v1.3.0
 	github.com/oracle/oci-go-sdk/v65 v65.128.0
 	github.com/ovn-kubernetes/libovsdb v0.9.0
@@ -161,7 +162,6 @@ require (
 	github.com/muesli/termenv v0.16.0 // indirect
 	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
 	github.com/nats-io/jwt/v2 v2.8.2 // indirect
-	github.com/nats-io/nkeys v0.4.16 // indirect
 	github.com/nats-io/nuid v1.0.1 // indirect
 	github.com/opencontainers/go-digest v1.0.0 // indirect
 	github.com/opencontainers/image-spec v1.1.1 // indirect
@@ -77,8 +77,8 @@ func TestEC2ErrorCatalogMetadataAndClasses(t *testing.T) {
 	if catalog.metadata.VerifiedOn == "" {
 		t.Fatal("verifiedOn is empty")
 	}
-	if len(catalog.codes) != 114 {
-		t.Fatalf("catalog code count = %d, want 114", len(catalog.codes))
+	if len(catalog.codes) != 115 {
+		t.Fatalf("catalog code count = %d, want 115", len(catalog.codes))
 	}
 	if catalog.codes["InvalidParameterValue"] != ec2ErrorClassClient {
 		t.Fatal("InvalidParameterValue is not classified as client")
@@ -5,6 +5,6 @@ package awsmodel
 // ModelCommit is the github.com/aws/api-models-aws commit the embedded
 // models were taken from, and ModelCommitDate the day it was committed.
 const (
-	ModelCommit     = "9b1a4929274443d8671ea8acfda7afbe856e29e9"
-	ModelCommitDate = "2026-09-25"
+	ModelCommit     = "e10d6a1c5ac36af7c0c61d1f67b3c3ebbe41de49"
+	ModelCommitDate = "2026-10-28"
 )
@@ -21,13 +21,13 @@ func TestLoadAllModels(t *testing.T) {
 		shapes     int
 	}{
 		ACM:                    {"2015-12-08", 40, 239},
-		EC2:                    {"2016-11-15", 804, 4245},
-		ECR:                    {"2015-09-21", 58, 404},
-		ECS:                    {"2014-11-13", 77, 643},
-		EKS:                    {"2017-11-01", 70, 450},
+		EC2:                    {"2016-11-15", 811, 4297},
+		ECR:                    {"2015-09-21", 62, 426},
+		ECS:                    {"2014-11-13", 88, 734},
+		EKS:                    {"2017-11-01", 70, 458},
 		ElasticLoadBalancingV2: {"2015-12-01", 51, 418},
 		IAM:                    {"2010-05-08", 180, 624},
-		RDS:                    {"2014-10-31", 164, 791},
+		RDS:                    {"2014-10-31", 168, 816},
 		Route53:                {"2013-04-01", 71, 409},
 		S3:                     {"2006-03-01", 112, 710},
 		STS:                    {"2011-06-15", 11, 101},
@@ -50,6 +50,7 @@
    "InvalidCapacityReservationId.Malformed",
    "InvalidCapacityReservationId.NotFound",
    "InvalidEgressOnlyInternetGatewayId.NotFound",
+    "InvalidGatewayID.NotFound",
    "InvalidGroup.Duplicate",
    "InvalidGroup.NotFound",
    "InvalidGroup.Reserved",
@@ -111,7 +111,7 @@ func GenerateRequests(service Service, operationName string, options RequestOpti
 	}

 	for _, member := range append([]string{""}, optional...) {
-		if unguessable[member] {
+		if unguessable[member] && options.Fixtures[operationName+"."+member] == "" && options.Fixtures[member] == "" {
 			continue
 		}
 		value, _, err := generate()
@@ -157,7 +157,8 @@ func GenerateRequests(service Service, operationName string, options RequestOpti

 // unguessable members take values the model cannot describe: a pagination
 // token the service issued, or a filter name only its documentation lists.
-// They get rejection requests but no acceptance request.
+// They get rejection requests, and an acceptance request only when a fixture
+// supplies a value.
 var unguessable = map[string]bool{
 	"Filter":    true,
 	"Filters":   true,
@@ -337,6 +337,20 @@ func TestGenerateRequestsSkipsAcceptanceForUnguessableMembers(t *testing.T) {
 	require.NotEmpty(t, markerRejections, "a pagination token still gets rejection requests")
 }

+func TestGenerateRequestsAcceptsAFixtureForAnUnguessableMember(t *testing.T) {
+	options := testRequestOptions
+	options.Fixtures = map[string]string{"ListRoles.Marker": "fixture-marker"}
+	plan, err := GenerateRequests(IAM, "ListRoles", options)
+	require.NoError(t, err)
+	var markers []any
+	for _, request := range plan.Cases {
+		if request.Acceptance() && request.Member == "Marker" {
+			markers = append(markers, request.Input["Marker"])
+		}
+	}
+	require.Equal(t, []any{"fixture-marker"}, markers)
+}
+
 func TestGenerateRequestsUsesResourceHints(t *testing.T) {
 	tests := []struct {
 		service   Service
@@ -7,9 +7,10 @@ Node installation, dev-environment lifecycle, guest image builds, and verificati
 | Script | What it does |
 |---|---|
 | `setup.sh` | The production binary installer, served from `install.mulgadc.com`. Installs deps, users, sudoers, the host firewall policy, files, systemd units, logrotate and udev rules, and loads `openvswitch`, `geneve`, `vhost_net` and `tun`, listing them in `/etc/modules-load.d/spinifex.conf` for later boots. It masks firewalld wherever it is installed, since the Spinifex nftables policy is the only supported node firewall. An unset `--firewall` keeps the mode recorded in `/etc/spinifex/firewall/mode`, and otherwise arms the policy when firewalld was active. Where systemd-resolved is installed but not running, resolvconf is absent and NetworkManager runs, it hands host DNS to resolved: it saves `/etc/resolv.conf` once to `/var/lib/spinifex-reset/resolv.conf.orig`, links `/etc/resolv.conf` to the resolved stub, and stops unless resolved has upstream servers and the package repository host still resolves. It links `/usr/bin/spx` to `/usr/local/bin/spx` where sudo's PATH leaves out `/usr/local/bin`, and takes the nbdkit plugin directory from `nbdkit --dump-config`, stopping when nbdkit cannot say. `SETUP_STAGES` runs a subset; `ISO_BUILD=1` runs it inside the ISO builder's chroot. |
-| `setup-ovn.sh` | Bootstraps a compute node for OVN VPC networking: installs OVN/OVS, creates `br-int`, configures the WAN bridge (bridged, DHCP, or routed-NAT uplink), sets the chassis identity, and applies overlay sysctls. It runs OVS and OVN as root, setting `OVS_USER_ID` and `OVN_USER_ID` where `/etc/sysconfig` holds them, and hooks the socket-permission helper onto `ovs-vswitchd`, `ovn-controller`, `openvswitch-ipsec` and, on EL, `ovn-northd`. `--management` also starts the OVN central services, writes their `ovn-ctl` flags to `/etc/default/ovn-central` (Debian) or `/etc/sysconfig/ovn-northd` (EL), and records the role in `/etc/spinifex/ovn-role`. A later run without `--management` refuses while that file, or the `OVN_CTL_OPTS` line older nodes carry in `/etc/default/ovn-central`, records the role. It adds `maxsize 100M` to the OVS and OVN logrotate files, plus `rotate 5` where the file sets no count, and rotates as root, matching the daemons. On a host whose IPsec stack is libreswan it stops and masks `openvswitch-ipsec.service` and `ipsec.service`, because OVN IPsec does not run on libreswan yet; elsewhere it enables and starts `openvswitch-ipsec.service`, unless the daemon has masked it to keep IPsec off. The links it creates persist through systemd-networkd where that is enabled, and otherwise through `/etc/spinifex/hostnet.json`, which `spinifex-hostnet.service` (`spx admin hostnet apply`) replays at boot and on every OVS start, starting `ovn-controller` again when an OVS stop took it down; the first run records the choice in `/etc/spinifex/hostnet-backend`. Where NetworkManager runs, it leaves the links Spinifex owns unmanaged. |
+| `setup-ovn.sh` | Bootstraps a compute node for OVN VPC networking: installs OVN/OVS, creates `br-int`, configures the WAN bridge (bridged, DHCP, or routed-NAT uplink), sets the chassis identity, and applies overlay sysctls. It runs OVS and OVN as root, setting `OVS_USER_ID` and `OVN_USER_ID` where `/etc/sysconfig` holds them, and hooks the socket-permission helper onto `ovs-vswitchd`, `ovn-controller`, `openvswitch-ipsec` and, on EL, `ovn-northd`. `--management` also starts the OVN central services, writes their `ovn-ctl` flags to `/etc/default/ovn-central` (Debian) or `/etc/sysconfig/ovn-northd` (EL), and records the role in `/etc/spinifex/ovn-role`. With clustered databases it returns only once this node's NB and SB members are back in their RAFT clusters under a known leader, and fails after `OVN_RAFT_REJOIN_WAIT` seconds (default 60) otherwise, so reruns one node at a time never take two members of a cluster down. A later run without `--management` refuses while that file, or the `OVN_CTL_OPTS` line older nodes carry in `/etc/default/ovn-central`, records the role. It adds `maxsize 100M` to the OVS and OVN logrotate files, plus `rotate 5` where the file sets no count, and rotates as root, matching the daemons. On a host whose IPsec stack is libreswan it enables, stops and masks `openvswitch-ipsec.service` and `ipsec.service` for the daemon to unmask when IPsec is on, unless `openvswitch-ipsec.service` is already running; elsewhere it enables and starts `openvswitch-ipsec.service`, unless the daemon has masked it to keep IPsec off. The links it creates persist through systemd-networkd where that is enabled, and otherwise through `/etc/spinifex/hostnet.json`, which `spinifex-hostnet.service` (`spx admin hostnet apply`) replays at boot and on every OVS start, starting `ovn-controller` again when an OVS stop took it down; the first run records the choice in `/etc/spinifex/hostnet-backend`. Where NetworkManager runs, it leaves the links Spinifex owns unmanaged. |
 | `install-node.sh` | Forms a multi-node cluster across hosts that are already running Spinifex: rebuilds the OVN database in clustered form, runs `spx admin init` on the first host and `spx admin join` on the rest, then verifies. Before touching any host it refuses hosts of different OS families (`distro_os_family`) or `spx` versions, and `--ipsec on` when any host uses libreswan. The scripted form of [install-multi-node](../docs/install/install-multi-node/README.md). Destroys the joining nodes' logical network state and crypto — see its `--help`. |
 | `dev-install.sh` | Full local dev setup through the production installer: builds from source, packs a tarball, runs `setup.sh`, initialises the cluster, and starts services under systemd. |
+| `build-oracle-appliance.sh` | Turns an Oracle Linux 10.2 system reachable over SSH with sudo into a sealed gold image, then powers it off. It updates OL and reboots into the newest UEK R8, sets SELinux permissive, installs Spinifex without initialising it, and leaves `spinifex.target` and the NATS watchdog disabled. It creates the `spinifex` login with a default password that must be changed at first login, installs `appliance/firstboot.sh` as `spinifex-appliance-firstboot.service`, and wipes the system's identity (machine-id, SSH host keys, OVS chassis, NetworkManager profiles, logs). On each unit's first boot that service puts the first physical NIC with carrier on a DHCP `br-wan`, names the host `spinifex-<mac>`, runs `setup-ovn.sh --management` and `spx admin init --nodes 1`, and shows the IP and login on the console and in `/etc/motd`. `--channel`, `--version` and `--tarball` pick what to install. |
 | `clone-deps.sh` | Clones or updates the viperblock and predastore checkouts alongside spinifex for cross-repo development. |

 ### `lib/distro.sh`: host-OS capability probes
@@ -58,7 +59,7 @@ Both ship to `/usr/local/share/spinifex/` on every node, so they are available o

 | Script | What it does |
 |---|---|
-| `node-reset.sh` | Clears **state** and leaves the node installed: stops services, destroys guests, wipes `/etc/spinifex`, `/var/lib/spinifex`, `/var/log/spinifex` and the OVN databases, and removes the OVS bridges. It stops OVS and OVN central by the unit names the host's packages use. It leaves the NetworkManager unmanaged-device config, which the next `setup-ovn.sh` run rewrites, so NetworkManager cannot take a NIC back in between. The node is then ready for a fresh `spx admin init`. `--keep-data` preserves `/var/lib/spinifex`; `SPX_RESET_CONFIRM=reset` runs it unattended. |
+| `node-reset.sh` | Clears **state** and leaves the node installed: stops services, destroys guests, wipes `/etc/spinifex`, `/var/lib/spinifex`, `/var/log/spinifex` and the OVN databases, and removes the OVS bridges. On a libreswan host it deletes the `ovs_cert_*` and `ovs_certkey_*` entries `ovs-monitor-ipsec` imported into libreswan's NSS database, leaving other identities and `/etc/ipsec.d/*.conf` alone. It stops OVS and OVN central by the unit names the host's packages use. It leaves the NetworkManager unmanaged-device config, which the next `setup-ovn.sh` run rewrites, so NetworkManager cannot take a NIC back in between. The node is then ready for a fresh `spx admin init`. `--keep-data` preserves `/var/lib/spinifex`; `SPX_RESET_CONFIRM=reset` runs it unattended. |
 | `uninstall-spx.sh` | Removes the **software** — the inverse of `setup.sh`. Delegates the state teardown above to `node-reset.sh` rather than reimplementing it, then removes the binary, the nbdkit plugin, the systemd units and slices, the service users, the sudoers grant, the sysctl/logrotate/chrony/udev drop-ins, the `systemd-networkd` units, the host firewall policy and its loaded nft table, the Spinifex systemd-resolved drop-in, and the Spinifex CA from the host trust store. It unmasks firewalld and leaves it disabled. Where `setup.sh` handed DNS to systemd-resolved, it restores the saved `/etc/resolv.conf` and turns resolved off. Verifies afterwards and exits non-zero naming anything that survived, then lists the host settings `setup.sh` changed that it leaves in place, such as the dnf repositories, `OVS_USER_ID`, masked IPsec units and the SELinux mode. |

 `node-reset.sh` finds guests by process name, `qemu-system-*` or `qemu-kvm`, owned by the daemon's service user, and waits for them to exit before it deletes any state.
@@ -0,0 +1,146 @@
+#!/bin/bash
+# Spinifex appliance first boot, for Oracle Linux units cloned from the gold
+# image that build-oracle-appliance.sh seals. The OL counterpart of the ISO's
+# spinifex-firstboot.sh: it also does the per-unit work the ISO installer does
+# before first boot (hostname, br-wan on the NIC that is present), then sets
+# up a single node. On success it writes $DONE_MARKER and the unit's
+# ConditionPathExists=! stops it running again. A failed run leaves no marker,
+# so the next boot retries from the top; every step below is safe to repeat.
+set -euo pipefail
+
+DONE_MARKER=/var/lib/spinifex/.appliance-firstboot-done
+SHARE=/usr/local/share/spinifex
+
+# shellcheck source=lib.sh
+. "$SHARE/appliance/lib.sh"
+# shellcheck source=../lib/distro.sh
+. "$SHARE/lib/distro.sh"
+
+log() { echo "[appliance-firstboot] $*"; }
+
+if [ -f "$DONE_MARKER" ]; then
+    log "already complete, skipping"
+    exit 0
+fi
+
+# ── Primary NIC ─────────────────────────────────────────────────────────────
+# Links start down when NetworkManager has no profile for them, so bring every
+# physical NIC up to read its carrier, and give a cable up to 60s to come up.
+nics=$(appliance_physical_nics)
+if [ -z "$nics" ]; then
+    log "no physical Ethernet device found"
+    exit 1
+fi
+for nic in $nics; do
+    ip link set "$nic" up || true
+done
+for _i in $(seq 1 60); do
+    for nic in $nics; do
+        [ "$(cat "/sys/class/net/$nic/carrier" 2>/dev/null)" = 1 ] && break 2
+    done
+    sleep 1
+done
+NIC=$(appliance_pick_nic)
+MAC=$(cat "/sys/class/net/$NIC/address")
+log "primary NIC $NIC ($MAC), carrier $(cat "/sys/class/net/$NIC/carrier" 2>/dev/null || echo 0)"
+
+HOSTNAME_NEW=$(appliance_hostname "$MAC")
+hostnamectl set-hostname "$HOSTNAME_NEW"
+log "hostname $HOSTNAME_NEW"
+
+# ── br-wan on the primary NIC, DHCP ─────────────────────────────────────────
+# The bridge takes the NIC's MAC so a DHCP reservation made for the unit's
+# port keeps working. bridge.stp no: with STP on, the port sits in its
+# forwarding delay long enough for DHCP on the bridge to time out.
+for con in $(nmcli -g UUID,DEVICE connection show | awk -F: -v d="$NIC" '$2 == d {print $1}'); do
+    nmcli connection delete "$con" || true
+done
+nmcli connection delete br-wan br-wan-port >/dev/null 2>&1 || true
+nmcli connection add type bridge ifname br-wan con-name br-wan \
+    bridge.stp no bridge.mac-address "$MAC" \
+    ipv4.method auto ipv6.method disabled connection.autoconnect yes
+nmcli connection add type bridge-slave ifname "$NIC" master br-wan con-name br-wan-port \
+    connection.autoconnect yes
+nmcli connection up br-wan || true
+
+log "waiting for a DHCP address on br-wan"
+WAN_IP=""
+for _i in $(seq 1 120); do
+    WAN_IP=$(ip -4 -o addr show br-wan scope global 2>/dev/null | awk '{print $4}' | cut -d/ -f1 | head -1)
+    if [ -n "$WAN_IP" ] && ip -4 route show default dev br-wan | grep -q .; then
+        break
+    fi
+    WAN_IP=""
+    sleep 1
+done
+if [ -z "$WAN_IP" ]; then
+    log "br-wan got no DHCP address and default route in 120s; check the cable on $NIC and the site's DHCP"
+    exit 1
+fi
+log "br-wan $WAN_IP"
+
+# ── Clock ───────────────────────────────────────────────────────────────────
+# Credentials and TLS certificates are minted below, and a certificate issued
+# against a wrong RTC carries a NotBefore that outlives the correction.
+# Bounded and non-fatal, as on the ISO.
+chronyc -a makestep >/dev/null 2>&1 || true
+if chronyc waitsync 12 0.1 0 5 >/dev/null 2>&1; then
+    log "clock synced: $(date -uIs)"
+else
+    log "WARNING: clock not synced, continuing with $(date -uIs)"
+fi
+
+# ── OVS and OVN ─────────────────────────────────────────────────────────────
+# Start them before setup-ovn.sh so their databases exist when it runs: on
+# physical hardware first-boot DB initialisation outlasts its own timeout.
+OVS_UNIT=$(distro_ovs_unit)
+OVN_LAYOUT=$(distro_ovn_central_layout)
+case $OVN_LAYOUT in
+    split) OVN_CENTRAL_UNIT=ovn-central.service ;;
+    lone-northd) OVN_CENTRAL_UNIT=ovn-northd.service ;;
+esac
+systemctl start "$OVS_UNIT" "$OVN_CENTRAL_UNIT"
+for _i in $(seq 1 120); do
+    if ovn-nbctl --timeout=2 get-connection >/dev/null 2>&1; then
+        log "OVN NB DB ready (${_i}s)"
+        break
+    fi
+    sleep 1
+done
+
+"$SHARE/setup-ovn.sh" --management
+
+# ── Single-node cluster ─────────────────────────────────────────────────────
+# Same arguments as the ISO's single-NIC DHCP path. spx writes credentials to
+# disk and never prints them.
+spx admin init --node "$HOSTNAME_NEW" --nodes 1 2>&1
+
+# spx ran as root, so the files it wrote are root:root. setup.sh's fixown
+# stage knows each file's owner and mode.
+SETUP_STAGES=fixown "$SHARE/setup.sh"
+
+# ── spinifex login ──────────────────────────────────────────────────────────
+if [ ! -f /home/spinifex/.ssh/id_ed25519 ]; then
+    install -d -m 700 -o spinifex -g spinifex /home/spinifex/.ssh
+    ssh-keygen -q -t ed25519 -f /home/spinifex/.ssh/id_ed25519 -N "" -C "spinifex@$HOSTNAME_NEW"
+    chown spinifex:spinifex /home/spinifex/.ssh/id_ed25519 /home/spinifex/.ssh/id_ed25519.pub
+fi
+# init wrote the AWS profile under HOME=/root; the operator logs in as spinifex.
+if [ -f /root/.aws/credentials ]; then
+    install -d -m 700 -o spinifex -g spinifex /home/spinifex/.aws
+    install -m 600 -o spinifex -g spinifex /root/.aws/credentials /home/spinifex/.aws/credentials
+    [ -f /root/.aws/config ] && install -m 600 -o spinifex -g spinifex /root/.aws/config /home/spinifex/.aws/config
+fi
+
+# ── Services ────────────────────────────────────────────────────────────────
+# Enabled here, not in the image, so a clone runs nothing of Spinifex before
+# init has written its config.
+systemctl enable spinifex.target spinifex-banner.service spinifex-firewall.service
+systemctl enable --now spinifex-nats-watchdog.timer
+
+mkdir -p "$(dirname "$DONE_MARKER")"
+touch "$DONE_MARKER"
+
+systemctl start spinifex.target
+/usr/local/bin/spx admin banner --boot-check
+log "complete: https://$WAN_IP:3000, login spinifex"
@@ -0,0 +1,4 @@
+# Holds the console login prompt until first boot and the banner have written
+# /etc/issue, so the unit's IP shows before anyone logs in.
+[Unit]
+After=spinifex-appliance-firstboot.service spinifex-banner.service
@@ -0,0 +1,37 @@
+#!/bin/bash
+# Helpers shared by the appliance first-boot script and its test. Sourced, not run.
+
+# appliance_physical_nics — physical Ethernet devices, one per line, in PCI
+# order. Skips anything without a backing device (bridges, veths, OVS ports,
+# loopback), non-Ethernet link types and wireless. SYSFS overrides /sys.
+appliance_physical_nics() {
+    local sys="${SYSFS:-/sys}/class/net" dev
+    for dev in "$sys"/*; do
+        [ -e "$dev/device" ] || continue
+        [ "$(cat "$dev/type" 2>/dev/null)" = 1 ] || continue
+        [ -d "$dev/wireless" ] && continue
+        printf '%s %s\n' "$(readlink -f "$dev/device")" "$(basename "$dev")"
+    done | sort | awk '{print $2}'
+}
+
+# appliance_pick_nic — the first physical NIC with carrier, else the first
+# physical NIC. Prints nothing when the host has none.
+appliance_pick_nic() {
+    local sys="${SYSFS:-/sys}/class/net" nic first=""
+    for nic in $(appliance_physical_nics); do
+        [ -n "$first" ] || first=$nic
+        if [ "$(cat "$sys/$nic/carrier" 2>/dev/null)" = 1 ]; then
+            echo "$nic"
+            return 0
+        fi
+    done
+    [ -n "$first" ] && echo "$first"
+    return 0
+}
+
+# appliance_hostname <mac> — spinifex-<last 6 hex digits of the MAC>.
+appliance_hostname() {
+    local hex
+    hex=$(echo "$1" | tr -d ':' | tr '[:upper:]' '[:lower:]')
+    echo "spinifex-${hex: -6}"
+}
@@ -0,0 +1,63 @@
+#!/bin/sh
+# lib_test.sh — tests for the appliance first-boot NIC and hostname helpers,
+# run against a fixture sysfs tree (SYSFS) so the host's own NICs never count.
+set -eu
+
+DIR=$(CDPATH='' cd -- "$(dirname -- "$0")" && pwd)
+TMP=$(mktemp -d)
+trap 'rm -rf "$TMP"' EXIT
+
+fails=0
+pass() { echo "  ok   - $1"; }
+fail() { echo "  FAIL - $1"; fails=$((fails + 1)); }
+
+echo "-- appliance/lib.sh"
+
+# nic <name> <pci-slot> <carrier> [type] — a physical NIC behind a PCI device.
+nic() {
+    mkdir -p "$SYS/devices/pci0000:00/$2" "$SYS/class/net/$1"
+    ln -s "../../../devices/pci0000:00/$2" "$SYS/class/net/$1/device"
+    echo "${4:-1}" > "$SYS/class/net/$1/type"
+    echo "$3" > "$SYS/class/net/$1/carrier"
+}
+# virt <name> — a link with no backing device (bridge, veth, OVS port).
+virt() {
+    mkdir -p "$SYS/class/net/$1"
+    echo 1 > "$SYS/class/net/$1/type"
+    echo 1 > "$SYS/class/net/$1/carrier"
+}
+fixture() { SYS="$TMP/$1"; mkdir -p "$SYS/class/net"; }
+run() { SYSFS="$SYS" bash -c ". '$DIR/lib.sh'; $1"; }
+
+check() {
+    if [ "$2" = "$3" ]; then pass "$1"; else fail "$1: want '$2', got '$3'"; fi
+}
+
+fixture order
+nic enp3s0 0000:00:03.0 0
+nic enp1s0 0000:00:01.0 0
+virt br-wan
+virt lo
+check "physical NICs in PCI order, virtual links skipped" "enp1s0 enp3s0" "$(run appliance_physical_nics | tr '\n' ' ' | sed 's/ $//')"
+check "no carrier anywhere picks the first by PCI order" enp1s0 "$(run appliance_pick_nic)"
+
+fixture carrier
+nic eno1 0000:00:01.0 0
+nic eno2 0000:00:02.0 1
+nic eno3 0000:00:03.0 1
+check "the first NIC with carrier wins over PCI order" eno2 "$(run appliance_pick_nic)"
+
+fixture wireless
+nic wlp1s0 0000:00:01.0 1
+mkdir -p "$SYS/class/net/wlp1s0/wireless"
+nic ib0 0000:00:02.0 1 32
+nic enp4s0 0000:00:04.0 0
+check "wireless and non-Ethernet links are skipped" enp4s0 "$(run appliance_pick_nic)"
+
+fixture none
+virt br-wan
+check "no physical NIC prints nothing" "" "$(run appliance_pick_nic)"
+
+check "hostname from the MAC's last 6 hex digits" spinifex-c0ffee "$(run 'appliance_hostname 52:54:00:C0:FF:EE')"
+
+[ "$fails" -eq 0 ] || { echo "  $fails failed"; exit 1; }
@@ -0,0 +1,191 @@
+#!/bin/bash
+# Runs as root on the Oracle Linux build system, one stage per call, driven by
+# build-oracle-appliance.sh from the payload directory it copied over.
+#   prepare.sh check      refuse anything but OL 10.2+ on UEK R8, x86_64
+#   prepare.sh update     dnf update, SELinux permissive; prints PREPARE_REBOOT=1 if
+#                         the newest kernel is not the running one
+#   prepare.sh install    install Spinifex, uninitialised and not enabled
+#   prepare.sh appliance  spinifex login, first-boot service, console placeholder
+#   prepare.sh seal       wipe this system's identity, then power off
+# INSTALL_SPINIFEX_CHANNEL, INSTALL_SPINIFEX_VERSION and INSTALL_SPINIFEX_TARBALL
+# pick the release tarball to install. BUILD_USER is the SSH user to lock at seal.
+set -euo pipefail
+
+# Shared by every unit built from this image; first login forces a change.
+DEFAULT_PASSWORD=spinifex
+
+HERE=$(cd "$(dirname "$0")" && pwd)
+SHARE=/usr/local/share/spinifex
+
+log() { echo "[prepare] $*"; }
+die() { echo "[prepare] ERROR: $*" >&2; exit 1; }
+
+[ "$(id -u)" = 0 ] || die "run as root"
+
+stage_check() {
+    # shellcheck source=/dev/null
+    . /etc/os-release
+    [ "${ID:-}" = ol ] || die "not Oracle Linux (ID=${ID:-unset})"
+    local major=${VERSION_ID%%.*} minor=${VERSION_ID#*.}
+    [ "$minor" = "$VERSION_ID" ] && minor=0
+    if [ "$major" -lt 10 ] || { [ "$major" -eq 10 ] && [ "$minor" -lt 2 ]; }; then
+        # A 10.0 or 10.1 system reaches 10.2 through the update stage, which
+        # calls this again with "strict" once it has run.
+        if [ "$major" -ne 10 ] || [ "${1:-}" = strict ]; then
+            die "Oracle Linux $VERSION_ID; 10.2 or later is required"
+        fi
+        log "Oracle Linux $VERSION_ID; the update stage brings it to the current minor"
+    fi
+    [ "$(uname -m)" = x86_64 ] || die "$(uname -m); only x86_64 is supported"
+    case $(uname -r) in
+        6.12.0-*uek*) ;;
+        *) die "running kernel $(uname -r) is not UEK R8; install kernel-uek and boot it first" ;;
+    esac
+    command -v nmcli >/dev/null || die "NetworkManager (nmcli) is required"
+    log "OK: $PRETTY_NAME, kernel $(uname -r)"
+}
+
+stage_update() {
+    setenforce 0 2>/dev/null || true
+    sed -i 's/^SELINUX=.*/SELINUX=permissive/' /etc/selinux/config
+    [ "$(getenforce)" != Enforcing ] || die "SELinux is still enforcing"
+    dnf update -y
+    stage_check strict >/dev/null
+    local newest
+    newest=$(grubby --default-kernel | sed 's|^/boot/vmlinuz-||')
+    case $newest in
+        6.12.0-*uek*) ;;
+        *) die "the default boot kernel is $newest, not UEK R8; set a kernel-uek default with grubby" ;;
+    esac
+    if [ "$newest" != "$(uname -r)" ]; then
+        echo PREPARE_REBOOT=1
+    fi
+}
+
+stage_install() {
+    [ ! -e /etc/spinifex/spinifex.toml ] || die "/etc/spinifex/spinifex.toml exists; this system has been initialised, start from a clean OL install"
+    # install.mulgadc.com serves the latest stable setup.sh, which may predate
+    # Oracle Linux support, so run the setup.sh inside the release tarball.
+    local dir tarball
+    dir=$(mktemp -d)
+    if [ -n "${INSTALL_SPINIFEX_TARBALL:-}" ]; then
+        tarball=$INSTALL_SPINIFEX_TARBALL
+    else
+        local url="https://install.mulgadc.com/download/${INSTALL_SPINIFEX_VERSION:-${INSTALL_SPINIFEX_CHANNEL:-latest}}/amd64"
+        tarball=$dir/spinifex.tar.gz
+        log "downloading $url"
+        curl -fsSL -o "$tarball" "$url"
+        curl -fsSL -o "$dir/checksum.sha256" "$url.sha256"
+        [ "$(awk '{print $1}' "$dir/checksum.sha256")" = "$(sha256sum "$tarball" | awk '{print $1}')" ] \
+            || die "checksum mismatch for $url"
+    fi
+    tar -xzf "$tarball" -C "$dir" ./setup.sh
+    INSTALL_SPINIFEX_TARBALL=$tarball INSTALL_SPINIFEX_SKIP_NEWGRP=1 bash "$dir/setup.sh"
+    rm -rf "$dir"
+    # The installer enables these on a live host. The ISO image ships them off
+    # and first boot enables them after init; a clone that started them first
+    # would run every service with no config.
+    systemctl disable --now spinifex.target spinifex-nats-watchdog.timer
+    log "Spinifex $(spx version 2>/dev/null | head -1) installed, not initialised"
+}
+
+stage_appliance() {
+    # As on the ISO: the existing spinifex system group as primary, so the
+    # login can reach the sockets setup.sh group-owns, and passwordless sudo.
+    if ! id spinifex >/dev/null 2>&1; then
+        useradd --create-home --shell /bin/bash --gid spinifex --groups wheel spinifex
+    fi
+    echo "spinifex:$DEFAULT_PASSWORD" | chpasswd
+    chage -d 0 spinifex
+    echo 'spinifex ALL=(ALL) NOPASSWD:ALL' > /etc/sudoers.d/spinifex-admin
+    chmod 0440 /etc/sudoers.d/spinifex-admin
+    visudo -cf /etc/sudoers.d/spinifex-admin >/dev/null
+
+    # OL's pwquality gives one try per login, then ends the session, so say the
+    # rules up front. spx admin banner keeps /etc/motd content above its own
+    # section, and agetty shows /etc/issue.d beside the banner in /etc/issue.
+    local hint="First login: the new password needs 8+ characters and must not contain 'spinifex'."
+    printf '%s\n' "$hint" > /etc/motd
+    install -d /etc/issue.d
+    printf '%s\n\n' "$hint" > /etc/issue.d/spinifex-password.issue
+
+    # sshd takes the first value it reads, so this sorts ahead of the
+    # 50-cloud-init.conf that templates ship with passwords off.
+    printf 'PasswordAuthentication yes\n' > /etc/ssh/sshd_config.d/10-spinifex-appliance.conf
+
+    # No DHCP on every port: first boot puts the primary NIC on br-wan and
+    # leaves the rest alone, so no second port competes for the default route.
+    printf '[main]\nno-auto-default=*\n' > /etc/NetworkManager/conf.d/90-spinifex-appliance.conf
+
+    install -D -m 0644 "$HERE/lib.sh" "$SHARE/appliance/lib.sh"
+    install -D -m 0755 "$HERE/firstboot.sh" "$SHARE/appliance/firstboot.sh"
+    install -m 0644 "$HERE/spinifex-appliance-firstboot.service" "$HERE/spinifex-banner.service" /etc/systemd/system/
+    local getty
+    for getty in getty@tty1 serial-getty@ttyS0; do
+        install -D -m 0644 "$HERE/getty-wait.conf" "/etc/systemd/system/$getty.service.d/spinifex-wait.conf"
+    done
+    systemctl daemon-reload
+    systemctl enable spinifex-appliance-firstboot.service
+
+    # Cloud-init would re-run on each unit with no datasource and rewrite the
+    # network and users first boot owns.
+    if [ -d /etc/cloud ]; then
+        touch /etc/cloud/cloud-init.disabled
+    fi
+    log "spinifex login (default password '$DEFAULT_PASSWORD', change forced) and first-boot service installed"
+}
+
+stage_seal() {
+    dnf clean all
+
+    systemctl stop "$(. "$SHARE/lib/distro.sh" && distro_ovs_unit)" ovn-northd.service ovn-controller.service 2>/dev/null || true
+    # OVS writes the chassis UUID on first start; a shared one makes every unit
+    # the same chassis in OVN.
+    rm -f /etc/openvswitch/system-id.conf /etc/openvswitch/conf.db /etc/openvswitch/.conf.db.~lock~
+    rm -f /var/lib/ovn/*.db /etc/ovn/*.db
+    rm -f /etc/spinifex/node.conf
+
+    # Profiles for the build system's NICs, its DHCP leases and NM's
+    # per-machine secret. First boot writes br-wan for the unit's own NIC.
+    rm -f /etc/NetworkManager/system-connections/*
+    rm -f /var/lib/NetworkManager/*.lease /var/lib/NetworkManager/secret_key \
+        /var/lib/NetworkManager/timestamps /var/lib/NetworkManager/seen-bssids
+    if command -v cloud-init >/dev/null; then
+        cloud-init clean --logs || true
+    fi
+
+    # sshd-keygen regenerates host keys before sshd on the next boot; systemd
+    # generates a machine-id when the file is empty.
+    rm -f /etc/ssh/ssh_host_*
+    truncate -s 0 /etc/machine-id
+    rm -f /var/lib/dbus/machine-id /var/lib/systemd/random-seed
+    echo localhost > /etc/hostname
+
+    # The build user's access goes; the spinifex login is the way in.
+    if [ -n "${BUILD_USER:-}" ] && [ "$BUILD_USER" != spinifex ] && [ "$BUILD_USER" != root ]; then
+        local home
+        home=$(getent passwd "$BUILD_USER" | cut -d: -f6)
+        rm -f "$home/.ssh/authorized_keys" "$home/.bash_history"
+        passwd -l "$BUILD_USER" >/dev/null
+        rm -f /etc/sudoers.d/90-cloud-init-users /etc/sudoers.d/90-spinifex-build
+    fi
+    rm -f /root/.ssh/authorized_keys /root/.bash_history /home/spinifex/.bash_history
+
+    printf '\nSpinifex first boot is in progress. This screen updates when it finishes.\n\n' > /etc/issue
+    sed -i '/^# --- Spinifex ---$/,$d' /etc/motd 2>/dev/null || true
+
+    find /var/log -type f -exec truncate -s 0 {} + 2>/dev/null || true
+    rm -rf /var/log/journal/* /tmp/* /var/tmp/*
+    sync
+    log "sealed; powering off"
+    systemctl poweroff --no-block
+}
+
+case "${1:-}" in
+    check) stage_check ;;
+    update) stage_update ;;
+    install) stage_install ;;
+    appliance) stage_appliance ;;
+    seal) stage_seal ;;
+    *) die "usage: $0 check|update|install|appliance|seal" ;;
+esac
@@ -0,0 +1,22 @@
+[Unit]
+Description=Spinifex appliance first-boot provisioning
+After=NetworkManager.service network-online.target chronyd.service
+Wants=NetworkManager.service network-online.target
+ConditionPathExists=!/var/lib/spinifex/.appliance-firstboot-done
+
+[Service]
+Type=oneshot
+Environment=HOME=/root
+ExecStart=/usr/local/share/spinifex/appliance/firstboot.sh
+RemainAfterExit=yes
+# The banner is never enabled when first boot fails, so without this the
+# console would keep the "first boot in progress" placeholder.
+ExecStopPost=/bin/sh -c 'if [ "$EXIT_STATUS" != "0" ]; then printf "\n*** Spinifex first boot FAILED ***\nLog in as spinifex, then run:\n  sudo journalctl -u spinifex-appliance-firstboot.service\nCheck the network cable and DHCP, then reboot to retry.\n\n" > /etc/issue; fi'
+# setup-ovn.sh and spx admin init must not wedge multi-user.target and keep
+# getty from ever showing a login prompt.
+TimeoutStartSec=600s
+StandardOutput=journal+console
+StandardError=journal+console
+
+[Install]
+WantedBy=multi-user.target
@@ -0,0 +1,15 @@
+[Unit]
+Description=Spinifex console banner and boot health check
+After=network-online.target
+
+[Service]
+Type=oneshot
+ExecStart=/usr/local/bin/spx admin banner --boot-check
+RemainAfterExit=yes
+# Cap it so a stuck boot-check cannot hold getty behind the console drop-in.
+TimeoutStartSec=30s
+StandardOutput=journal
+StandardError=journal
+
+[Install]
+WantedBy=multi-user.target
@@ -0,0 +1,113 @@
+#!/bin/bash
+# build-oracle-appliance.sh — turn an Oracle Linux 10.2 system reachable over
+# SSH with sudo into a sealed Spinifex gold image: installed, not initialised,
+# with a first-boot service that sets each cloned unit up as a single node.
+# The system powers off at the end; capture its disk as the image.
+#
+# Usage: build-oracle-appliance.sh [options] <user>@<host>
+#   --channel latest|dev   install channel (default: latest)
+#   --version <tag>        install a specific release
+#   --tarball <path>       install a local release tarball instead
+#
+# Ports, keys and jump hosts come from ~/.ssh/config. <user> needs sudo;
+# without passwordless sudo, sudo asks for its password once.
+set -euo pipefail
+
+HERE=$(cd "$(dirname "$0")" && pwd)
+PAYLOAD_SRC="$HERE/appliance"
+REMOTE_DIR=/tmp/spinifex-appliance
+
+CHANNEL="" VERSION="" TARBALL="" TARGET=""
+
+usage() { sed -n '2,13p' "$0" | sed 's/^# \{0,1\}//'; exit "${1:-0}"; }
+log() { echo "==> $*"; }
+die() { echo "ERROR: $*" >&2; exit 1; }
+
+while [ $# -gt 0 ]; do
+    case "$1" in
+        --channel) CHANNEL="${2:-}"; shift ;;
+        --version) VERSION="${2:-}"; shift ;;
+        --tarball) TARBALL="${2:-}"; shift ;;
+        -h | --help) usage ;;
+        -*) die "unknown option: $1" ;;
+        *) [ -z "$TARGET" ] || die "one target only"; TARGET=$1 ;;
+    esac
+    shift
+done
+[ -n "$TARGET" ] || usage 1
+case $TARGET in *@*) ;; *) die "target must be <user>@<host>" ;; esac
+BUILD_USER=${TARGET%@*}
+[ -z "$TARBALL" ] || [ -f "$TARBALL" ] || die "no such tarball: $TARBALL"
+
+# One multiplexed connection; host key checking is relaxed because the build
+# system is usually a fresh install whose key nobody has recorded yet.
+CTL_DIR=$(mktemp -d)
+trap 'ssh -o ControlPath="$CTL_DIR/c" -O exit "$TARGET" >/dev/null 2>&1 || true; rm -rf "$CTL_DIR"' EXIT
+SSH_OPTS=(-o ControlMaster=auto -o ControlPath="$CTL_DIR/c" -o ControlPersist=60
+    -o StrictHostKeyChecking=accept-new -o ServerAliveInterval=15 -o ConnectTimeout=10)
+
+rsh() { ssh "${SSH_OPTS[@]}" "$TARGET" "$@"; }
+
+# prepare <stage> — run one prepare.sh stage as root on the build system.
+prepare() {
+    rsh "sudo env INSTALL_SPINIFEX_CHANNEL='$CHANNEL' INSTALL_SPINIFEX_VERSION='$VERSION' \
+        INSTALL_SPINIFEX_TARBALL='${TARBALL:+$REMOTE_DIR/$(basename "$TARBALL")}' \
+        BUILD_USER='$BUILD_USER' bash $REMOTE_DIR/prepare.sh $1"
+}
+
+push_payload() {
+    rsh "rm -rf $REMOTE_DIR && mkdir -p $REMOTE_DIR"
+    tar -C "$PAYLOAD_SRC" -cf - . | rsh "tar -C $REMOTE_DIR -xf -"
+    if [ -n "$TARBALL" ]; then
+        scp "${SSH_OPTS[@]}" -q "$TARBALL" "$TARGET:$REMOTE_DIR/"
+    fi
+}
+
+wait_for_ssh() {
+    local _i
+    ssh "${SSH_OPTS[@]}" -O exit "$TARGET" >/dev/null 2>&1 || true
+    sleep 10
+    for _i in $(seq 1 60); do
+        if rsh true 2>/dev/null; then return 0; fi
+        sleep 5
+    done
+    die "$TARGET did not come back over SSH after the reboot"
+}
+
+rsh true || die "cannot SSH to $TARGET; set its key in ~/.ssh/config (IdentityFile) or ssh-add it"
+# The stages run sudo with no terminal, so grant passwordless sudo for the
+# build when the user lacks it. Seal removes the grant.
+if ! rsh "sudo -n true" 2>/dev/null; then
+    log "Granting $BUILD_USER passwordless sudo for the build; sudo asks for its password once"
+    ssh "${SSH_OPTS[@]}" -t "$TARGET" "echo '$BUILD_USER ALL=(ALL) NOPASSWD:ALL' | sudo tee /etc/sudoers.d/90-spinifex-build >/dev/null && sudo chmod 0440 /etc/sudoers.d/90-spinifex-build" \
+        || die "could not grant $BUILD_USER passwordless sudo on the build system"
+    rsh "sudo -n true" || die "$BUILD_USER still has no passwordless sudo on the build system"
+fi
+# A minimal OL install has no tar, which the payload copy and the installer need.
+rsh "command -v tar >/dev/null || sudo dnf install -y -q tar" || die "cannot install tar on the build system"
+
+log "Copying the payload to $TARGET"
+push_payload
+prepare check
+
+log "Updating Oracle Linux"
+prepare update | tee "$CTL_DIR/update.log"
+if grep -qx 'PREPARE_REBOOT=1' "$CTL_DIR/update.log"; then
+    log "Rebooting into the updated kernel"
+    rsh "sudo systemctl reboot" || true
+    wait_for_ssh
+    # /tmp is cleared on boot.
+    push_payload
+    prepare check
+fi
+
+log "Installing Spinifex (not initialising it)"
+prepare install
+
+log "Installing the spinifex login and the first-boot service"
+prepare appliance
+
+log "Sealing; the system powers off when done"
+prepare seal || true
+
+log "Done. Capture $TARGET's disk once it has powered off; each unit set up from it initialises itself on first boot."
@@ -0,0 +1,475 @@
+#!/bin/bash
+# build-sqs-store-microvm-image.sh — build the SQS store microVM kernel +
+# initramfs: the dedicated nats-server, ext4 support, the store's guest steps
+# and a clean power-down. The guest never formats a disk; build/microvm/
+# sqs-store/sqs-store-lib.sh states its fw_cfg contract. The LB image
+# (build-microvm-image.sh) is separate and unchanged.
+#
+# Inputs:
+#   $REPO_ROOT/build/microvm/sqs-store/init.sh, inittab
+#   $REPO_ROOT/build/microvm/sqs-store/sqs-store-{lib.sh,prepare,run,stop}
+#   $REPO_ROOT/bin/nats-server  (make bin/nats-server, the go.mod-pinned version)
+#
+# Outputs:
+#   $REPO_ROOT/build/microvm/sqs-store/vmlinuz
+#   $REPO_ROOT/build/microvm/sqs-store/initramfs.cpio.gz
+#
+# Rootfs strategy (first match wins):
+#   1. apk available on host → install directly into a temp chroot (Alpine CI/CD)
+#   2. docker available      → run alpine container, export filesystem
+#   3. podman available      → same via podman
+set -euo pipefail
+
+REPO_ROOT="$(cd "$(dirname "$0")/.." && pwd)"
+BUILD_DIR="$REPO_ROOT/build/microvm/sqs-store"
+INIT_SH="$BUILD_DIR/init.sh"
+INITTAB="$BUILD_DIR/inittab"
+NATS_SERVER_BIN="$REPO_ROOT/bin/nats-server"
+
+ALPINE_VERSION="${ALPINE_VERSION:-3.24}"
+ALPINE_MIRROR="${ALPINE_MIRROR:-https://dl-cdn.alpinelinux.org/alpine}"
+# e2fsprogs: mkfs.ext4/e2fsck for a guest that mounts an ext4 data volume.
+# kmod: the image's own depmod/modprobe rebuild module metadata below.
+ALPINE_PACKAGES="busybox linux-virt iproute2 ca-certificates e2fsprogs kmod"
+# Modules kept by name on top of the virtio/fw_cfg/failover globs, each with its
+# full depends= and softdep closure. tiny_power_button turns the ACPI power
+# button (QMP system_powerdown) into a signal busybox init handles as poweroff.
+MICROVM_EXTRA_MODULES="${MICROVM_EXTRA_MODULES:-ext4 tiny_power_button}"
+
+echo "[build-sqs-store-microvm-image] repo root: $REPO_ROOT"
+echo "[build-sqs-store-microvm-image] build dir: $BUILD_DIR"
+mkdir -p "$BUILD_DIR"
+
+# --- Verify non-negotiable tools ---
+# fakeroot is mandatory: device nodes are written into the cpio archive inside a
+# faked-root environment so no real device node is ever created on the host
+# filesystem (which previously required sudo mknod and risked clobbering host
+# /dev/null on cleanup).
+for tool in cpio gzip find fakeroot; do
+    if ! command -v "$tool" >/dev/null 2>&1; then
+        echo "ERROR: required tool not found: $tool" >&2
+        exit 1
+    fi
+done
+
+# --- Verify the dedicated nats-server is the go.mod-pinned build ---
+# A stale or host-installed binary would ship an unreviewed server version.
+NATS_SERVER_PIN=$(awk '$1 == "github.com/nats-io/nats-server/v2" { print $2; exit }' "$REPO_ROOT/go.mod")
+if [ -z "$NATS_SERVER_PIN" ]; then
+    echo "ERROR: no github.com/nats-io/nats-server/v2 requirement in $REPO_ROOT/go.mod" >&2
+    exit 1
+fi
+if [ ! -x "$NATS_SERVER_BIN" ]; then
+    echo "ERROR: $NATS_SERVER_BIN not found — run 'make bin/nats-server'" >&2
+    exit 1
+fi
+if ! file -b "$NATS_SERVER_BIN" | grep -q "statically linked"; then
+    echo "ERROR: $NATS_SERVER_BIN is not statically linked — it cannot run on the musl guest" >&2
+    exit 1
+fi
+nats_server_version=$("$NATS_SERVER_BIN" --version 2>&1 | awk '{ print $NF; exit }')
+if [ "$nats_server_version" != "$NATS_SERVER_PIN" ]; then
+    echo "ERROR: $NATS_SERVER_BIN reports $nats_server_version, go.mod pins $NATS_SERVER_PIN" >&2
+    exit 1
+fi
+echo "[build-sqs-store-microvm-image] nats-server: $nats_server_version (matches go.mod)"
+
+# --- Create chroot ---
+CHROOT_DIR=$(mktemp -d)
+# Guard: every destructive op below targets "$CHROOT_DIR/...". An empty or
+# unexpected value would aim rm -rf at the host (e.g. /dev). Refuse anything that
+# is not a freshly-created temp dir.
+case "${CHROOT_DIR:?mktemp -d returned empty}" in
+    /tmp/*|/var/tmp/*|"${TMPDIR:-/nonexistent}"/*) ;;
+    *) echo "ERROR: refusing to use unexpected CHROOT_DIR: $CHROOT_DIR" >&2; exit 1 ;;
+esac
+# Outputs are written here first, beside $BUILD_DIR so publishing is a rename.
+STAGE_DIR=$(mktemp -d "$BUILD_DIR/.stage.XXXXXX")
+CONTAINER_TOOL=""
+CONTAINER_CID=""
+cleanup() {
+    echo "[build-sqs-store-microvm-image] cleaning up chroot: $CHROOT_DIR"
+    rm -rf "$CHROOT_DIR"
+    rm -rf "${STAGE_DIR:?}"
+    if [ -n "$CONTAINER_CID" ] && [ -n "$CONTAINER_TOOL" ]; then
+        echo "[build-sqs-store-microvm-image] cleaning up container: $CONTAINER_CID"
+        $CONTAINER_TOOL rm -f "$CONTAINER_CID" >/dev/null 2>&1 || true
+    fi
+}
+trap cleanup EXIT
+
+# --- Populate rootfs ---
+# Strategy 1: native apk (Alpine hosts, CI)
+if command -v apk >/dev/null 2>&1; then
+    echo "[build-sqs-store-microvm-image] rootfs: native apk"
+    mkdir -p \
+        "$CHROOT_DIR/etc/apk" \
+        "$CHROOT_DIR/lib/apk/db" \
+        "$CHROOT_DIR/var/cache/apk"
+    cat > "$CHROOT_DIR/etc/apk/repositories" <<EOF
+${ALPINE_MIRROR}/v${ALPINE_VERSION}/main
+${ALPINE_MIRROR}/v${ALPINE_VERSION}/community
+EOF
+    apk add \
+        --root "$CHROOT_DIR" \
+        --arch x86_64 \
+        --no-cache \
+        --allow-untrusted \
+        --initdb \
+        $ALPINE_PACKAGES
+
+# Strategy 2: docker -> BuildKit local export.
+# Do NOT use `docker export` here. On Docker's overlay2 driver it leaks the
+# container's overlay mount (the mount survives `docker rm`), orphaning layers
+# under /var/lib/docker that only a daemon restart reclaims; on a busy CI host
+# this grows without bound. `buildx --output type=local` writes the image
+# filesystem straight to a directory and releases every mount, so nothing leaks.
+elif command -v docker >/dev/null 2>&1 && docker info >/dev/null 2>&1; then
+    echo "[build-sqs-store-microvm-image] rootfs: docker buildx --output type=local (alpine:${ALPINE_VERSION})"
+    BUILD_CTX=$(mktemp -d)
+    cat > "$BUILD_CTX/Dockerfile" <<DOCKERFILE
+FROM alpine:${ALPINE_VERSION}
+RUN apk add --no-cache ${ALPINE_PACKAGES}
+DOCKERFILE
+    docker buildx build \
+        --platform linux/amd64 \
+        --output "type=local,dest=${CHROOT_DIR}" \
+        "$BUILD_CTX"
+    rm -rf "$BUILD_CTX"
+
+# Strategy 3: podman -> run + export. Podman's store does not exhibit the docker
+# overlay2 export-mount leak, and podman has no buildx.
+elif command -v podman >/dev/null 2>&1; then
+    echo "[build-sqs-store-microvm-image] rootfs: podman export (alpine:${ALPINE_VERSION})"
+    CONTAINER_TOOL=podman
+    CONTAINER_CID=$(podman run -d \
+        "alpine:${ALPINE_VERSION}" \
+        sh -c "apk add --no-cache ${ALPINE_PACKAGES}")
+
+    exit_code=$(podman wait "$CONTAINER_CID")
+    if [ "$exit_code" != "0" ]; then
+        echo "ERROR: package installation failed in container (exit $exit_code)" >&2
+        exit 1
+    fi
+
+    podman export "$CONTAINER_CID" | tar -x -C "$CHROOT_DIR"
+
+else
+    echo "ERROR: no rootfs build tool available." >&2
+    echo "       Install one of: apk (Alpine), docker, podman" >&2
+    exit 1
+fi
+
+# --- /dev: empty mountpoint only ---
+# tar extraction without root silently creates 0-byte regular files for device
+# nodes. Drop them and leave an empty /dev mountpoint. The two static nodes the
+# kernel needs (/dev/console to wire init's stdio, /dev/null) are written into
+# the cpio archive under fakeroot at pack time — NOT created on the host. init
+# mounts devtmpfs over /dev as its first action for everything else at runtime.
+rm -rf "${CHROOT_DIR:?}/dev"
+mkdir "$CHROOT_DIR/dev"
+
+# --- Place init script ---
+echo "[build-sqs-store-microvm-image] installing init.sh as /init..."
+if [ ! -f "$INIT_SH" ]; then
+    echo "ERROR: $INIT_SH not found — cannot build initramfs" >&2
+    exit 1
+fi
+install -m 0755 "$INIT_SH" "$CHROOT_DIR/init"
+
+# --- Place inittab ---
+# Required: without it busybox init runs its default table, not the power-down.
+if [ ! -f "$INITTAB" ]; then
+    echo "ERROR: $INITTAB not found — cannot build initramfs" >&2
+    exit 1
+fi
+echo "[build-sqs-store-microvm-image] installing inittab..."
+install -m 0644 "$INITTAB" "$CHROOT_DIR/etc/inittab"
+
+# --- Place the store's guest steps ---
+echo "[build-sqs-store-microvm-image] installing store guest steps..."
+mkdir -p "$CHROOT_DIR/usr/local/lib/sqs-store"
+for step in sqs-store-lib.sh sqs-store-prepare sqs-store-run sqs-store-stop; do
+    if [ ! -f "$BUILD_DIR/$step" ]; then
+        echo "ERROR: $BUILD_DIR/$step not found — cannot build initramfs" >&2
+        exit 1
+    fi
+done
+install -m 0644 "$BUILD_DIR/sqs-store-lib.sh" "$CHROOT_DIR/usr/local/lib/sqs-store/"
+install -m 0755 "$BUILD_DIR/sqs-store-prepare" "$BUILD_DIR/sqs-store-run" "$BUILD_DIR/sqs-store-stop" \
+    "$CHROOT_DIR/usr/local/lib/sqs-store/"
+
+# --- Copy nats-server binary ---
+echo "[build-sqs-store-microvm-image] copying nats-server binary..."
+mkdir -p "$CHROOT_DIR/usr/local/bin"
+install -m 0755 "$NATS_SERVER_BIN" "$CHROOT_DIR/usr/local/bin/nats-server"
+
+# --- Create required directories for init ---
+mkdir -p \
+    "$CHROOT_DIR/proc" \
+    "$CHROOT_DIR/sys" \
+    "$CHROOT_DIR/dev" \
+    "$CHROOT_DIR/etc/ssl/certs"
+
+# --- Locate kernel ---
+KERNEL_IMG=$(find "$CHROOT_DIR/boot" -name "vmlinuz*" | sort | tail -1)
+if [ -z "$KERNEL_IMG" ]; then
+    echo "ERROR: no vmlinuz found in $CHROOT_DIR/boot" >&2
+    exit 1
+fi
+echo "[build-sqs-store-microvm-image] kernel image: $KERNEL_IMG"
+
+KERNEL_VER=$(basename "$KERNEL_IMG" | sed 's/vmlinuz-//')
+KERNEL_CONFIG="$CHROOT_DIR/boot/config-${KERNEL_VER}"
+
+# --- Assert fw_cfg module ---
+if ! find "$CHROOT_DIR/lib/modules" \
+        -name "qemu_fw_cfg.ko*" \
+        -o -name "fw_cfg_sysfs.ko*" \
+    | grep -q .; then
+    if ! grep -q "CONFIG_FW_CFG_SYSFS=y" "$KERNEL_CONFIG" 2>/dev/null; then
+        echo "ERROR: qemu_fw_cfg module missing from initramfs — fw_cfg reads will fail at boot" >&2
+        exit 1
+    fi
+    echo "[build-sqs-store-microvm-image] fw_cfg: built-in (CONFIG_FW_CFG_SYSFS=y)"
+else
+    echo "[build-sqs-store-microvm-image] fw_cfg: module found in lib/modules"
+fi
+
+# --- Assert ext4 userspace tools (data volume) ---
+for fs_tool in mkfs.ext4 e2fsck; do
+    if [ ! -x "$CHROOT_DIR/sbin/$fs_tool" ] && [ ! -x "$CHROOT_DIR/usr/sbin/$fs_tool" ]; then
+        echo "ERROR: $fs_tool missing from image — guest cannot check or create an ext4 data volume" >&2
+        exit 1
+    fi
+done
+echo "[build-sqs-store-microvm-image] e2fsprogs: mkfs.ext4 + e2fsck present"
+
+# --- Strip kernel modules (Phase 2 keep-list approach) ---
+# Only modules needed by a virtio-mmio microvm guest are kept:
+#   virtio*.ko*    — virtio bus, net, blk, rng, console, and helpers
+#   *fw_cfg*.ko*   — QEMU fw_cfg sysfs driver (delivers boot blobs)
+#   *failover*.ko* — net_failover + failover (virtio_net deps for live migration)
+#   $MICROVM_EXTRA_MODULES — ext4 by default, plus dependency closure
+# All other loadable modules are removed. Drivers compiled into the kernel
+# (CONFIG_VIRTIO_MMIO=y etc.) are unaffected — they live in vmlinuz, not here.
+
+# module_path prints a module's kernel/ path from the stock modules.dep, by
+# name or modules.alias; it fails when the module is not loadable.
+module_path() {
+    local kdir="$1" name="${2//-/_}" path alias_target
+    path=$(awk -F: -v n="$name" '{ f = $1; sub(/.*\//, "", f); sub(/\.ko(\.gz|\.xz|\.zst)?$/, "", f);
+        gsub(/-/, "_", f); if (f == n) { print $1; exit } }' "$kdir/modules.dep")
+    if [ -z "$path" ] && [ -f "$kdir/modules.alias" ]; then
+        alias_target=$(awk -v n="$2" '$1 == "alias" && $2 == n { print $3; exit }' "$kdir/modules.alias")
+        [ -n "$alias_target" ] && [ "$alias_target" != "$2" ] && path=$(module_path "$kdir" "$alias_target")
+    fi
+    [ -n "$path" ] && printf '%s\n' "$path"
+}
+
+# module_builtin reports whether a module name is compiled into the kernel.
+module_builtin() {
+    local kdir="$1" name="${2//-/_}"
+    [ -f "$kdir/modules.builtin" ] || return 1
+    awk -v n="$name" '{ f = $0; sub(/.*\//, "", f); sub(/\.ko$/, "", f); gsub(/-/, "_", f);
+        if (f == n) { found = 1; exit } } END { exit !found }' "$kdir/modules.builtin"
+}
+
+# image_kmod runs the image's own kmod (depmod, modprobe) through its musl
+# loader, so module metadata matches the guest's modprobe without a host kmod.
+image_kmod() {
+    "$CHROOT_DIR/lib/ld-musl-x86_64.so.1" --library-path "$CHROOT_DIR/lib:$CHROOT_DIR/usr/lib" \
+        --argv0 "$1" "$CHROOT_DIR/bin/kmod" "${@:2}"
+}
+
+echo "[build-sqs-store-microvm-image] stripping kernel modules to virtio+fw_cfg+${MICROVM_EXTRA_MODULES} only..."
+for kver_dir in "$CHROOT_DIR/lib/modules"/*/; do
+    kver="$(basename "$kver_dir")"
+    kernel_dir="$kver_dir/kernel"
+    [ -d "$kernel_dir" ] || continue
+
+    if [ ! -s "$kver_dir/modules.dep" ]; then
+        echo "ERROR: stock ${kver_dir}modules.dep missing — cannot resolve module dependencies" >&2
+        exit 1
+    fi
+
+    # Extra modules start the closure; the glob families are kept as found.
+    keep_list=$(mktemp)
+    : > "$keep_list"
+    for name in $MICROVM_EXTRA_MODULES; do
+        if ! path=$(module_path "$kver_dir" "$name"); then
+            if module_builtin "$kver_dir" "$name"; then continue; fi
+            echo "ERROR: module $name is neither loadable nor built into $kver" >&2
+            exit 1
+        fi
+        echo "$path" >> "$keep_list"
+    done
+
+    # Close over modules.dep hard dependencies and modules.softdep pre:/post:
+    # names until the set stops growing. A softdep must be loadable or built in.
+    while :; do
+        before=$(sort -u "$keep_list" | wc -l)
+        for path in $(sort -u "$keep_list"); do
+            awk -F': ' -v p="$path" '$1 == p { n = split($2, d, " "); for (i = 1; i <= n; i++) print d[i] }' \
+                "$kver_dir/modules.dep" >> "$keep_list"
+            mod=$(basename "$path"); mod="${mod%%.ko*}"; mod="${mod//-/_}"
+            [ -f "$kver_dir/modules.softdep" ] || continue
+            for soft in $(awk -v n="$mod" '{ m = $2; gsub(/-/, "_", m) }
+                    $1 == "softdep" && m == n { for (i = 3; i <= NF; i++) if ($i != "pre:" && $i != "post:") print $i }' \
+                    "$kver_dir/modules.softdep"); do
+                if soft_path=$(module_path "$kver_dir" "$soft"); then
+                    echo "$soft_path" >> "$keep_list"
+                elif ! module_builtin "$kver_dir" "$soft"; then
+                    echo "ERROR: softdep $soft of $mod is neither loadable nor built into $kver" >&2
+                    exit 1
+                fi
+            done
+        done
+        after=$(sort -u "$keep_list" | wc -l)
+        [ "$before" -eq "$after" ] && break
+    done
+
+    extra_names=$(sort -u "$keep_list" | while read -r p; do m=$(basename "$p"); m="${m%%.ko*}"; echo "${m//-/_}"; done)
+    (cd "$kver_dir" && find kernel -name "virtio*.ko*" -o -name "*fw_cfg*.ko*" \
+        -o -name "*failover*.ko*") >> "$keep_list"
+
+    # Copy the keeper set out, preserving each kernel/ relative path.
+    tmp_keep=$(mktemp -d)
+    while read -r rel; do
+        rel="${rel#kernel/}"
+        mkdir -p "$tmp_keep/$(dirname "$rel")"
+        cp "$kernel_dir/$rel" "$tmp_keep/$rel"
+    done < <(sort -u "$keep_list")
+    rm -f "$keep_list"
+
+    # Replace kernel module tree with the keeper set.
+    rm -rf "$kernel_dir"
+    mkdir -p "$kernel_dir"
+    # Restore each kept file into the same relative path.
+    (cd "$tmp_keep" && find . -name "*.ko*" | while read -r f; do
+        dest_dir="$kernel_dir/$(dirname "${f#./}")"
+        mkdir -p "$dest_dir"
+        cp "$f" "$dest_dir/"
+    done)
+    rm -rf "$tmp_keep"
+
+    # Decompress kept modules to plain .ko. Alpine ships .ko.gz, but a host
+    # kmod that lacks gzip support (e.g. Debian trixie's kmod 34) silently
+    # produces an EMPTY modules.dep from compressed modules — the guest's
+    # modprobe then no-ops and qemu_fw_cfg never loads. Plain .ko sidesteps the
+    # host-kmod dependency entirely and the guest needs no decompressor either.
+    find "$kernel_dir" -name "*.ko.gz" -exec gunzip -f {} +
+
+    # Independent closure gate: each extra-closure module's own .modinfo depends=
+    # must name a kept module or a built-in, so a resolver bug fails here.
+    kept_names=$(find "$kernel_dir" -name "*.ko" -exec basename {} .ko \; | tr '-' '_' | sort -u)
+    while read -r ko; do
+        ko_name=$(basename "$ko" .ko); ko_name="${ko_name//-/_}"
+        grep -qx "$ko_name" <<< "$extra_names" || continue
+        deps=$(tr '\0' '\n' < "$ko" | grep -a -m1 '^depends=' | cut -d= -f2 | tr ',' ' ' || true)
+        for dep in $deps; do
+            dep="${dep//-/_}"
+            if ! grep -qx "$dep" <<< "$kept_names" && ! module_builtin "$kver_dir" "$dep"; then
+                echo "ERROR: $(basename "$ko") depends on $dep, which is neither kept nor built in" >&2
+                exit 1
+            fi
+        done
+    done < <(find "$kernel_dir" -name "*.ko")
+    for name in $MICROVM_EXTRA_MODULES; do
+        if ! grep -qx "${name//-/_}" <<< "$kept_names" && ! module_builtin "$kver_dir" "$name"; then
+            echo "ERROR: required module $name missing from the stripped module tree" >&2
+            exit 1
+        fi
+    done
+    echo "[build-sqs-store-microvm-image] extra module closure: $(echo "$extra_names" | tr '\n' ' ')"
+    echo "[build-sqs-store-microvm-image] modules kept: $(echo "$kept_names" | tr '\n' ' ')"
+
+    # Rebuild module metadata with the image's own kmod, the binary the guest's
+    # modprobe runs, so no host depmod is needed. Stock indexes still name .ko.gz
+    # files, so they are removed first and a failed rebuild cannot ship them.
+    for f in dep dep.bin alias alias.bin softdep weakdep symbols symbols.bin \
+        builtin.bin builtin.alias.bin devname; do
+        rm -f "$kver_dir/modules.$f"
+    done
+    if ! image_kmod depmod -b "$CHROOT_DIR" "$kver"; then
+        echo "ERROR: image depmod failed for $kver" >&2
+        exit 1
+    fi
+    if [ ! -s "$kver_dir/modules.dep" ] || [ ! -s "$kver_dir/modules.dep.bin" ]; then
+        echo "ERROR: ${kver_dir}modules.dep or modules.dep.bin missing or empty after depmod" >&2
+        exit 1
+    fi
+    # Resolution gate: the image's modprobe must resolve every required module
+    # to plain .ko files present in the stripped tree.
+    for name in $MICROVM_EXTRA_MODULES qemu_fw_cfg virtio_net; do
+        module_builtin "$kver_dir" "$name" && continue
+        if ! plan=$(image_kmod modprobe -d "$CHROOT_DIR" -S "$kver" --show-depends "$name"); then
+            echo "ERROR: image modprobe cannot resolve $name for $kver" >&2
+            exit 1
+        fi
+        while read -r verb path _; do
+            [ "$verb" = "insmod" ] || continue
+            case "$path" in
+                "$CHROOT_DIR/lib/modules/$kver/kernel/"*.ko) [ -f "$path" ] && continue ;;
+            esac
+            echo "ERROR: modprobe plan for $name names $path, not a kept plain .ko" >&2
+            exit 1
+        done <<< "$plan"
+        echo "[build-sqs-store-microvm-image] modprobe $name: $(awk '{ n = split($2, a, "/"); printf "%s ", a[n] }' <<< "$plan")"
+    done
+done
+
+# --- Strip package-manager artifacts (not needed at runtime) ---
+echo "[build-sqs-store-microvm-image] stripping package manager artifacts..."
+rm -rf \
+    "$CHROOT_DIR/lib/apk" \
+    "$CHROOT_DIR/var/cache/apk" \
+    "$CHROOT_DIR/etc/apk" \
+    "$CHROOT_DIR/usr/share/man" \
+    "$CHROOT_DIR/usr/share/doc" \
+    "$CHROOT_DIR/usr/share/apk" \
+    "$CHROOT_DIR/usr/include" \
+    "$CHROOT_DIR/usr/lib/pkgconfig"
+
+# --- Strip debug symbols from binaries ---
+# nats-server in /usr/local/bin is left as built (-s -w), so the shipped bytes
+# are the ones whose version was checked above.
+echo "[build-sqs-store-microvm-image] stripping debug symbols from binaries..."
+find "$CHROOT_DIR/usr/sbin" "$CHROOT_DIR/usr/bin" \
+     "$CHROOT_DIR/sbin" "$CHROOT_DIR/bin" \
+    -type f 2>/dev/null | while read -r bin; do
+    strip --strip-all "$bin" 2>/dev/null || true
+done
+
+# --- Copy vmlinuz out before stripping /boot from chroot ---
+VMLINUZ_OUT="$STAGE_DIR/vmlinuz"
+echo "[build-sqs-store-microvm-image] copying kernel: $VMLINUZ_OUT"
+cp "$KERNEL_IMG" "$VMLINUZ_OUT"
+
+# Drop /boot from the chroot — vmlinuz, Alpine's stock initramfs-virt, System.map
+# and kernel config all live here (~25 MiB). None are needed at runtime: the
+# guest already has the kernel loaded by QEMU's -kernel, and Alpine's initramfs
+# is unused because we replace it with our own /init. Leaving /boot in the cpio
+# nearly doubles uncompressed initramfs size and forces a 512 MiB guest just to
+# survive decompression.
+rm -rf "${CHROOT_DIR:?}/boot"
+
+# --- Build initramfs ---
+INITRAMFS_OUT="$STAGE_DIR/initramfs.cpio.gz"
+echo "[build-sqs-store-microvm-image] building initramfs: $INITRAMFS_OUT"
+# Create the static device nodes and pack the archive inside a single fakeroot
+# session: mknod/cpio see faked device entries and write real char-device
+# records into the cpio, while the host filesystem never gets an actual node.
+fakeroot sh -c '
+    cd "$1" || exit 1
+    mknod -m 600 dev/console c 5 1
+    mknod -m 666 dev/null    c 1 3
+    find . | cpio --quiet -o -H newc | gzip -9 > "$2"
+' _ "$CHROOT_DIR" "$INITRAMFS_OUT"
+
+# --- Gate the staged pair, then publish it ---
+# Nothing reaches $BUILD_DIR until every gate passes, so a failed build cannot
+# leave a pair the grouped make target would take as up to date.
+"$REPO_ROOT/scripts/lib/sqs-store-microvm-publish.sh" "$STAGE_DIR" "$BUILD_DIR"
+echo ""
+echo "[build-sqs-store-microvm-image] done."
@@ -298,7 +298,6 @@ log "preflight on $N host(s)"

 versions=""
 families=""
-libreswan_hosts=""
 for i in $(seq 0 $((N - 1))); do
    host="${HOSTS[$i]}"

@@ -322,12 +321,6 @@ for i in $(seq 0 $((N - 1))); do
        fail "$host: cannot tell its OS from /etc/os-release"
    families+="${host}: ${family}"$'\n'

-    # libreswan's pluto: Spinifex cannot run OVN IPsec on it yet, and spx admin
-    # init refuses --ipsec=true there, which would stop formation part way.
-    if [ "$(out "$host" "test -x /usr/libexec/ipsec/pluto && echo yes")" = "yes" ]; then
-        libreswan_hosts+=" $host"
-    fi
-
    # A node that is already in a multi-node cluster is one whose keys and
    # logical network state are load-bearing. Refusing here is deliberate:
    # forming a cluster must never be the command that destroys a live one.
@@ -386,12 +379,6 @@ $(sed '/^$/d; s/^/         /' <<<"$families")
       A cluster must run one OS on every host."
 fi

-if $IPSEC_ENABLED && [ -n "$libreswan_hosts" ]; then
-    fail "--ipsec on, but these hosts use libreswan for IPsec:$libreswan_hosts
-       Spinifex cannot run OVN IPsec on libreswan yet. Re-run with --ipsec off
-       only if the nodes share a trusted private link."
-fi
-
 # --- Confirm ---------------------------------------------------------------

 echo ""
@@ -460,7 +447,7 @@ done
 # Matched on the service argv so the ssh shell running this cannot match itself.
 for host in "${HOSTS[@]}"; do
    $DRY_RUN && { echo "  [dry-run] $host: wait for every 'spx service' process to exit"; continue; }
-    on "$host" "for u in spinifex-predastore spinifex-viperblock spinifex-sqs-store spinifex-nats; do
+    on "$host" "for u in spinifex-predastore spinifex-viperblock spinifex-nats; do
            systemctl is-active --quiet \$u && sudo systemctl stop \$u
        done
        for _ in \$(seq 1 45); do
@@ -0,0 +1,158 @@
+#!/bin/sh
+# ipsec_state_helper_test.sh — tests for the plaintext Geneve guard in the
+# IPsec state helper that setup.sh installs.
+#
+# The helper is cut out of setup.sh's heredoc and run against stub nft,
+# ovs-vsctl and systemctl commands, with pluto's path pointed at a fixture.
+set -eu
+
+SCRIPT_DIR=$(CDPATH='' cd -- "$(dirname -- "$0")" && pwd)
+SETUP="$SCRIPT_DIR/setup.sh"
+TMP=$(mktemp -d)
+trap 'rm -rf "$TMP"' EXIT
+
+fails=0
+pass() { echo "  ok   - $1"; }
+fail() { echo "  FAIL - $1"; fails=$((fails + 1)); }
+
+echo "-- IPsec state helper: plaintext Geneve guard"
+
+HELPER="$TMP/helper"
+awk '/tee "\$\{IPSEC_STATE_HELPER\}"/ { on = 1; next } on && /^HELPER$/ { exit } on { print }' "$SETUP" |
+    sed "s#/usr/libexec/ipsec/pluto#$TMP/pluto#g" > "$HELPER"
+chmod +x "$HELPER"
+grep -q '^    guard)' "$HELPER" || { echo "could not extract the helper from setup.sh" >&2; exit 1; }
+
+mkdir -p "$TMP/bin"
+cat > "$TMP/bin/nft" <<EOF
+#!/bin/sh
+[ "\$*" = "-f -" ] || { echo "nft \$*" >> "$TMP/nft.log"; exit 0; }
+cat >> "$TMP/nft.log"
+echo "--" >> "$TMP/nft.log"
+EOF
+cat > "$TMP/bin/ovs-vsctl" <<EOF
+#!/bin/sh
+[ -e "$TMP/ovs-down" ] && { echo "ovs-vsctl: database connection failed" >&2; exit 1; }
+case "\$*" in
+    *" find Interface type=geneve") cat "$TMP/ifaces" ;;
+    *" get Interface "*)
+        for last; do :; done
+        f="$TMP/opt.\$5.\${last#options:}"
+        if [ -e "\$f" ]; then cat "\$f"; else echo; fi
+        ;;
+    *" get Open_vSwitch . external_ids:ovn-encap-ip")
+        if [ -e "$TMP/encap" ]; then cat "$TMP/encap"; else echo; fi
+        ;;
+    *) exit 1 ;;
+esac
+EOF
+printf '#!/bin/sh\nexit 0\n' > "$TMP/bin/systemctl"
+chmod +x "$TMP/bin/nft" "$TMP/bin/ovs-vsctl" "$TMP/bin/systemctl"
+
+# reset — no tunnels, OVS up, no pluto.
+reset() {
+    rm -f "$TMP"/opt.* "$TMP/ovs-down" "$TMP/nft.log" "$TMP/pluto" "$TMP/encap"
+    : > "$TMP/ifaces"
+    touch "$TMP/nft.log"
+}
+# tunnel <name> <local_ip|-> [remote_name] — a geneve interface as ovn-controller writes it.
+tunnel() {
+    echo "$1" >> "$TMP/ifaces"
+    [ "$2" = - ] || printf '"%s"\n' "$2" > "$TMP/opt.$1.local_ip"
+    [ -z "${3:-}" ] || echo "$3" > "$TMP/opt.$1.remote_name"
+}
+run() {
+    RC=0
+    OUT=$(PATH="$TMP/bin:$PATH" "$HELPER" "$1" 2>&1) || RC=$?
+    NFTLOG=$(cat "$TMP/nft.log")
+}
+LIFTED="table inet spinifex_ipsec_guard
+delete table inet spinifex_ipsec_guard
+--"
+# expect_guarded <desc> <sources> — one transaction that swaps in a table
+# dropping Geneve without an SA from exactly those sources.
+expect_guarded() {
+    if [ "$RC" = 0 ] &&
+        [ "$(grep -c '^--$' "$TMP/nft.log")" = 1 ] &&
+        [ "$(sed -n 2p "$TMP/nft.log")" = "delete table inet spinifex_ipsec_guard" ] &&
+        grep -q "^        elements = { $2 }$" "$TMP/nft.log" &&
+        grep -q '^        ip saddr @srcs udp dport 6081 rt ipsec missing drop$' "$TMP/nft.log"; then
+        pass "$1"
+    else
+        fail "$1 (rc=$RC): got [$NFTLOG] want sources [$2] $OUT"
+    fi
+}
+expect_lifted() {
+    if [ "$RC" = 0 ] && [ "$NFTLOG" = "$LIFTED" ]; then pass "$1"; else fail "$1 (rc=$RC): got [$NFTLOG] $OUT"; fi
+}
+
+reset
+tunnel ovn-node2-0 10.26.7.1 node2
+tunnel ovn-node3-0 10.26.7.1 node3
+run guard
+expect_guarded "guards the encap address once OVN wants the tunnels encrypted" "10.26.7.1/32"
+
+reset
+tunnel ovn-node2-0 10.26.7.1
+tunnel ovn-node3-0 10.26.7.1
+run guard
+expect_lifted "lifts the guard while OVN runs the tunnels in plaintext"
+
+reset
+tunnel ovn-node2-0 10.26.7.1 node2
+tunnel ovn-node3-0 10.26.8.1
+run guard
+expect_guarded "guards only the addresses of tunnels OVN encrypts" "10.26.7.1/32"
+
+reset
+tunnel ovn-node2-0 - node2
+run guard
+expect_guarded "guards a tunnel without a local_ip from every source" "0.0.0.0/0"
+
+reset
+echo '"10.26.7.1"' > "$TMP/encap"
+run guard
+expect_guarded "guards the encap address before ovn-controller has built the tunnels" "10.26.7.1/32"
+
+reset
+echo '"10.26.7.1,10.26.8.1"' > "$TMP/encap"
+run guard
+expect_guarded "guards each encap address of a multi-address host before the tunnels exist" "10.26.7.1/32,10.26.8.1/32"
+
+reset
+run guard
+expect_guarded "guards every source before the tunnels exist when no encap address is set" "0.0.0.0/0"
+
+reset
+tunnel ovn-node2-0 10.26.7.1 node2
+touch "$TMP/ovs-down"
+run guard
+if [ "$RC" != 0 ] && [ -z "$NFTLOG" ]; then
+    pass "keeps the guard and fails when OVS cannot be read"
+else
+    fail "with OVS down: rc=$RC nft calls [$NFTLOG]"
+fi
+
+reset
+touch "$TMP/pluto"
+chmod +x "$TMP/pluto"
+run off
+expect_lifted "off lifts the guard on a libreswan host"
+
+reset
+run off
+if [ "$RC" = 0 ] && [ -z "$NFTLOG" ]; then pass "off leaves nft alone without libreswan"; else fail "off without pluto: rc=$RC [$NFTLOG]"; fi
+
+reset
+run unguard
+expect_lifted "unguard lifts the guard"
+
+reset
+run bogus
+if [ "$RC" = 2 ] && [ -z "$NFTLOG" ]; then pass "refuses an unknown verb"; else fail "unknown verb: rc=$RC [$NFTLOG]"; fi
+
+if [ "$fails" -gt 0 ]; then
+    echo "$fails failure(s)"
+    exit 1
+fi
+echo "all passed"
@@ -0,0 +1,288 @@
+#!/bin/sh
+# sqs-store-guest_test.sh — the SQS store guest steps start the store only on
+# a valid contract and one checked, mounted disk, never format, and stop the
+# store before unmounting it.
+#
+# Drives build/microvm/sqs-store/sqs-store-lib.sh against a fake fw_cfg, sysfs
+# and stub tools. The stop cases signal only test-owned children.
+# The case settings below are read by the sourced library, not by this file.
+# shellcheck disable=SC2034
+set -eu
+
+ROOT=$(CDPATH='' cd -- "$(dirname -- "$0")/../.." && pwd)
+LIB=${SQS_GUEST_LIB:-"$ROOT/build/microvm/sqs-store/sqs-store-lib.sh"}
+TMP=$(mktemp -d)
+CHILDREN=""
+cleanup() {
+    for p in $CHILDREN; do kill -KILL "$p" 2>/dev/null || true; done
+    rm -rf "$TMP"
+}
+trap cleanup EXIT
+
+fails=0
+pass() { echo "  ok   - $1"; }
+fail() { echo "  FAIL - $1"; fails=$((fails + 1)); }
+check() { if eval "$2"; then pass "$1"; else fail "$1"; fi; }
+
+echo "-- build/microvm/sqs-store/sqs-store-lib.sh"
+
+SERIAL=vol0123456789abcdef0
+NKEY=UAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAB
+ADDR=10.11.12.13
+
+# Stub tools record each call to calls.log; their exit codes come from files.
+mkdir -p "$TMP/bin"
+cat > "$TMP/bin/e2fsck" <<'EOF'
+#!/bin/sh
+echo "e2fsck $*" >> "$T/calls.log"
+exit "$(cat "$T/fsck.rc")"
+EOF
+cat > "$TMP/bin/mount" <<'EOF'
+#!/bin/sh
+echo "mount $*" >> "$T/calls.log"
+exit "$(cat "$T/mount.rc")"
+EOF
+cat > "$TMP/bin/umount" <<'EOF'
+#!/bin/sh
+echo "umount $*" >> "$T/calls.log"
+EOF
+for f in mkfs mkfs.ext4 mke2fs; do
+    printf '#!/bin/sh\necho "mkfs %s $*" >> "$T/calls.log"\n' "$f" > "$TMP/bin/$f"
+done
+cat > "$TMP/bin/ip" <<'EOF'
+#!/bin/sh
+cat "$T/ip.out"
+EOF
+cat > "$TMP/bin/nats-server" <<'EOF'
+#!/bin/sh
+echo "nats-server $*" >> "$T/calls.log"
+[ "$1" = -t ] && exit "$(cat "$T/configtest.rc")"
+exit 0
+EOF
+chmod +x "$TMP/bin"/*
+
+# fresh <name>: a case directory with a valid contract and one matching disk
+# among others. Callers then break one thing.
+fresh() {
+    T="$TMP/$1"
+    export T
+    mkdir -p "$T/fwcfg" "$T/sys" "$T/dev" "$T/run" "$T/data"
+    for k in data-serial listen-addr client-nkey; do mkdir -p "$T/fwcfg/$k"; done
+    printf '%s\n' "$SERIAL" > "$T/fwcfg/data-serial/raw"
+    printf '%s\n' "$ADDR" > "$T/fwcfg/listen-addr/raw"
+    printf '%s\n' "$NKEY" > "$T/fwcfg/client-nkey/raw"
+    disk vda vol00000000000000000
+    disk vdb "$SERIAL"
+    echo 0 > "$T/fsck.rc"
+    echo 0 > "$T/mount.rc"
+    echo 0 > "$T/configtest.rc"
+    : > "$T/calls.log"
+    : > "$T/console"
+    printf '1: lo    inet 127.0.0.1/8 scope host lo\n2: eth0    inet %s/24 brd 10.11.12.255 scope global eth0\n' "$ADDR" > "$T/ip.out"
+}
+disk() { mkdir -p "$T/sys/$1"; printf '%s\n' "$2" > "$T/sys/$1/serial"; : > "$T/dev/$1"; }
+
+# prepare: run sqs_prepare in a subshell with every path pointed at the case.
+prepare() {
+    (
+        SQS_FWCFG="$T/fwcfg" SQS_SYSBLOCK="$T/sys" SQS_DEVDIR="$T/dev" SQS_NODE_TEST=-e
+        SQS_RUN="$T/run" SQS_DATA="$T/data" SQS_CONSOLE="$T/console"
+        E2FSCK="$TMP/bin/e2fsck" MOUNT="$TMP/bin/mount" IP="$TMP/bin/ip" NATS_BIN="$TMP/bin/nats-server"
+        PATH="$TMP/bin:$PATH"
+        export PATH
+        # shellcheck source=../../build/microvm/sqs-store/sqs-store-lib.sh
+        . "$LIB"
+        sqs_prepare
+    ) > /dev/null 2>&1
+}
+refused() { [ ! -e "$T/run/ready" ] && grep -q FATAL "$T/console"; }
+never_mounted() { ! grep -q '^mount ' "$T/calls.log"; }
+never_formatted() { ! grep -q '^mkfs ' "$T/calls.log"; }
+never_checked() { ! grep -q '^e2fsck ' "$T/calls.log"; }
+
+# --- A valid contract prepares the store ---
+fresh valid
+check "valid contract is prepared" 'prepare && [ -e "$T/run/ready" ]'
+check "the matching disk is checked, then mounted at the data dir" \
+    'grep -qx "e2fsck -p $T/dev/vdb" "$T/calls.log" && grep -qx "mount -t ext4 -o nodev,nosuid,noexec $T/dev/vdb $T/data" "$T/calls.log"'
+check "config listens only on the supplied address" \
+    'grep -c "listen" "$T/run/nats.conf" | grep -qx 1 && grep -qx "listen: \"$ADDR:4222\"" "$T/run/nats.conf"'
+check "config admits only the supplied user NKey" 'grep -q "users = \[ { nkey: \"$NKEY\" } \]" "$T/run/nats.conf"'
+check "config syncs JetStream on the data disk" \
+    'grep -q "store_dir: \"$T/data\"" "$T/run/nats.conf" && grep -q "sync_interval: always" "$T/run/nats.conf"'
+check "config is tested by the server before ready" 'grep -qx "nats-server -t -c $T/run/nats.conf" "$T/calls.log"'
+check "no auth material on the console" '! grep -q "$NKEY" "$T/console"'
+check "prepare reports the disk and config, not a listener" 'grep -q "prepared: $T/dev/vdb mounted" "$T/console" && ! grep -qi "listening" "$T/console"'
+
+fresh fsck-corrected; echo 1 > "$T/fsck.rc"
+check "e2fsck exit 1 (corrected) proceeds" 'prepare && [ -e "$T/run/ready" ]'
+
+# --- Disk refusals: nothing mounted, nothing formatted ---
+fresh no-disk; rm -rf "$T/sys/vdb"
+check "missing disk is refused before any check or mount" '! prepare && refused && never_checked && never_mounted'
+
+fresh wrong-disk; printf 'vol0fffffffffffffff0\n' > "$T/fwcfg/data-serial/raw"
+check "disk with another serial is refused" '! prepare && refused && never_checked && never_mounted'
+
+fresh two-disks; disk vdc "$SERIAL"
+check "two disks with the serial are refused" '! prepare && refused && never_checked && never_mounted'
+
+fresh bad-serial; printf 'vol-0123456789abcdef\n' > "$T/fwcfg/data-serial/raw"
+check "malformed serial is refused" '! prepare && refused && never_mounted'
+
+fresh no-node; rm -f "$T/dev/vdb"
+check "disk without a device node is refused" '! prepare && refused && never_mounted'
+
+for rc in 2 4 8; do
+    fresh "fsck-$rc"; echo "$rc" > "$T/fsck.rc"
+    check "e2fsck exit $rc is refused without mount or format" '! prepare && refused && never_mounted && never_formatted'
+done
+
+fresh mount-fail; echo 32 > "$T/mount.rc"
+check "failed mount is refused" '! prepare && refused'
+
+# --- Auth and listener refusals happen before the disk is touched ---
+fresh no-nkey; rm -rf "$T/fwcfg/client-nkey"
+check "missing NKey is refused before the disk" '! prepare && refused && never_checked && never_mounted'
+
+for bad in seed lower short two empty; do
+    fresh "nkey-$bad"
+    case $bad in
+        seed) v=SUAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA ;;
+        lower) v=uaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaab ;;
+        short) v=UAAAA ;;
+        two) v="$NKEY
+$NKEY" ;;
+        empty) v="" ;;
+    esac
+    printf '%s\n' "$v" > "$T/fwcfg/client-nkey/raw"
+    check "NKey ($bad) is refused before the disk, without echoing it" \
+        '! prepare && refused && never_checked && { [ -z "$v" ] || ! grep -qF "$v" "$T/console"; }'
+done
+
+fresh nkey-checksum; echo 1 > "$T/configtest.rc"
+check "NKey the server rejects is refused before the disk" '! prepare && refused && never_checked && never_mounted'
+
+fresh no-addr; rm -rf "$T/fwcfg/listen-addr"
+check "missing listen address is refused" '! prepare && refused && never_mounted'
+
+fresh addr-absent; printf '1: lo    inet 127.0.0.1/8 scope host lo\n2: eth0    inet 10.11.12.99/24 scope global eth0\n' > "$T/ip.out"
+check "listen address not on a guest NIC is refused" '! prepare && refused && never_mounted'
+
+fresh addr-loopback; printf '127.0.0.1\n' > "$T/fwcfg/listen-addr/raw"
+check "loopback listen address is refused" '! prepare && refused && never_mounted'
+
+for bad in 0.0.0.0x 1.2.3 1.2.3.256 01.2.3.4; do
+    fresh "addr-$bad"; printf '%s\n' "$bad" > "$T/fwcfg/listen-addr/raw"
+    check "malformed listen address $bad is refused" '! prepare && refused && never_mounted'
+done
+
+# --- Run: the store starts only when prepared and not stopping ---
+run_step() {
+    (
+        SQS_RUN="$T/run" SQS_CONSOLE="$T/console" NATS_BIN="$TMP/bin/nats-server"
+        # shellcheck source=../../build/microvm/sqs-store/sqs-store-lib.sh
+        . "$LIB"
+        sqs_run
+    ) > /dev/null 2>&1
+}
+fresh run-ready; prepare; : > "$T/calls.log"
+check "a prepared store starts the server" 'run_step && grep -qx "nats-server -c $T/run/nats.conf" "$T/calls.log"'
+fresh run-unready
+check "an unprepared store never starts the server" '! run_step && ! grep -q "^nats-server" "$T/calls.log"'
+fresh run-stopping; prepare; : > "$T/calls.log"; : > "$T/run/stopping"
+check "a stopping store never starts the server" '! run_step && ! grep -q "^nats-server" "$T/calls.log"'
+
+# --- Stop: SIGTERM the verified store, then unmount ---
+STORE="$TMP/bin/fake-store"
+cp "$(command -v sleep)" "$STORE"
+stop_step() {
+    (
+        SQS_RUN="$T/run" SQS_DATA="$T/data" SQS_CONSOLE="$T/console" SQS_MOUNTS="$T/mounts"
+        UMOUNT="$TMP/bin/umount" NATS_BIN="$STORE" SQS_STOP_BUDGET_TENTHS=10
+        # shellcheck source=../../build/microvm/sqs-store/sqs-store-lib.sh
+        . "$LIB"
+        sqs_stop
+    ) > /dev/null 2>&1
+}
+gone() { [ ! -r "/proc/$1/status" ] || grep -q '^State:[[:space:]]*Z' "/proc/$1/status"; }
+# record <pid>: write the identity sqs_run would record for that process.
+# shellcheck source=../../build/microvm/sqs-store/sqs-store-lib.sh
+record() { echo "$1 $( . "$LIB"; sqs_starttime "$1" )" > "$T/run/pid"; }
+
+fresh stop-orderly; echo "dev $T/data ext4 rw 0 0" > "$T/mounts"
+"$STORE" 60 & pid=$!; CHILDREN="$CHILDREN $pid"; record "$pid"
+stop_step
+check "stop sends SIGTERM and the store exits" 'gone "$pid" && grep -q "SIGTERM sent" "$T/console" && ! grep -q SIGKILL "$T/console"'
+check "stop blocks respawn and unmounts after the store exits" \
+    '[ -e "$T/run/stopping" ] && grep -qx "umount $T/data" "$T/calls.log"'
+
+fresh stop-hung; echo "dev $T/data ext4 rw 0 0" > "$T/mounts"
+(trap '' TERM; exec "$STORE" 60) & pid=$!; CHILDREN="$CHILDREN $pid"; record "$pid"
+stop_step
+check "store ignoring SIGTERM is killed after the budget, then unmounted" \
+    'gone "$pid" && grep -q SIGKILL "$T/console" && grep -qx "umount $T/data" "$T/calls.log"'
+
+fresh stop-reused; echo "dev $T/data ext4 rw 0 0" > "$T/mounts"
+"$STORE" 60 & pid=$!; CHILDREN="$CHILDREN $pid"; echo "$pid 1" > "$T/run/pid"
+stop_step
+check "a recorded PID with another start time is never signalled" '! gone "$pid" && grep -q "no running store" "$T/console"'
+kill -KILL "$pid"
+
+# A start held after its stopping check and before its exec: the run step
+# blocks writing "run: starting" to a FIFO console until the test reads it.
+fresh stop-pending-start; echo "dev $T/data ext4 rw 0 0" > "$T/mounts"
+: > "$T/run/ready"
+mkfifo "$T/run-console"
+printf '#!/bin/sh\n: > "%s/started"\nexec "%s" 60\n' "$T" "$STORE" > "$TMP/bin/pending-nats"
+chmod +x "$TMP/bin/pending-nats"
+SQS_RUN="$T/run" SQS_CONSOLE="$T/run-console" NATS_BIN="$TMP/bin/pending-nats" \
+    sh -c '. "$1"; sqs_run' sh "$LIB" & pid=$!; CHILDREN="$CHILDREN $pid"
+i=0
+while [ "$i" -lt 100 ] && ! grep -q "^$pid " "$T/run/pid" 2>/dev/null; do i=$((i + 1)); sleep 0.05; done
+check "pending start is held before its exec" 'grep -q "^$pid " "$T/run/pid" && ! gone "$pid" && [ ! -e "$T/started" ]'
+stop_step
+# Release the FIFO: a start that survived the stop would exec the server now.
+cat "$T/run-console" > /dev/null 2>&1 &
+reader=$!; CHILDREN="$CHILDREN $reader"
+i=0
+while [ "$i" -lt 20 ] && ! gone "$pid"; do i=$((i + 1)); sleep 0.05; done
+check "stop ends a pending start before it can exec the server" 'gone "$pid" && [ ! -e "$T/started" ]'
+check "stop unmounts only after the pending start is gone" \
+    'grep -q "store exited" "$T/console" && grep -qx "umount $T/data" "$T/calls.log"'
+kill -KILL "$reader" 2>/dev/null || true
+
+# --- Shutdown: the generic remount runs only once the store is gone ---
+# shutdown_step [survivor-pid]: with a pid, that process is reported alive
+# whatever happens to it, as a writer stuck past SIGKILL would be.
+shutdown_step() {
+    (
+        SQS_RUN="$T/run" SQS_DATA="$T/data" SQS_CONSOLE="$T/console" SQS_MOUNTS="$T/mounts"
+        UMOUNT="$TMP/bin/umount" SQS_STOP_BUDGET_TENTHS=5
+        # shellcheck source=../../build/microvm/sqs-store/sqs-store-lib.sh
+        . "$LIB"
+        if [ -n "${1:-}" ]; then sqs_same() { [ "$1" = "$survivor" ]; }; survivor=$1; fi
+        sqs_shutdown
+    ) > /dev/null 2>&1
+}
+
+fresh shutdown-orderly; echo "dev $T/data ext4 rw 0 0" > "$T/mounts"
+"$STORE" 60 & pid=$!; CHILDREN="$CHILDREN $pid"; record "$pid"
+check "orderly shutdown succeeds" 'shutdown_step'
+check "orderly shutdown unmounts the data disk, then runs umount -a -r" \
+    '[ "$(grep "^umount" "$T/calls.log" | tr "\n" "|")" = "umount $T/data|umount -a -r|" ]'
+
+fresh shutdown-survivor; echo "dev $T/data ext4 rw 0 0" > "$T/mounts"
+"$STORE" 60 & pid=$!; CHILDREN="$CHILDREN $pid"; record "$pid"
+check "shutdown with a surviving store reports failure" '! shutdown_step "$pid"'
+check "a surviving store leaves the data disk and every other mount alone" \
+    '! grep -q "^umount" "$T/calls.log" && grep -q "STILL ALIVE" "$T/console" && grep -q "skipping umount -a -r" "$T/console"'
+
+fresh stop-unmounted; : > "$T/mounts"
+stop_step
+check "stop with no data mount does not unmount" '! grep -q "^umount" "$T/calls.log"'
+
+if [ "$fails" -ne 0 ]; then
+    echo "  $fails failure(s)"
+    exit 1
+fi
@@ -0,0 +1,97 @@
+#!/bin/bash
+# sqs-store-microvm-publish.sh STAGE_DIR OUT_DIR — gate a staged SQS store
+# microVM kernel + initramfs and move them into OUT_DIR only if every gate passes.
+#
+# A failed gate leaves OUT_DIR as it was: a previous valid pair stays, and an
+# older pair stays older than the build inputs, so make rebuilds next time.
+set -euo pipefail
+
+STAGE_DIR="${1:?usage: sqs-store-microvm-publish.sh STAGE_DIR OUT_DIR}"
+OUT_DIR="${2:?usage: sqs-store-microvm-publish.sh STAGE_DIR OUT_DIR}"
+VMLINUZ="$STAGE_DIR/vmlinuz"
+INITRAMFS="$STAGE_DIR/initramfs.cpio.gz"
+tag="[build-sqs-store-microvm-image]"
+
+for f in "$VMLINUZ" "$INITRAMFS"; do
+    if [ ! -s "$f" ]; then
+        echo "ERROR: staged artifact $f missing or empty" >&2
+        exit 1
+    fi
+done
+
+# --- Content gate: read the packed archive, not the chroot ---
+if ! archive_list=$(zcat "$INITRAMFS" | cpio --quiet -t); then
+    echo "ERROR: $INITRAMFS is not a readable gzip cpio archive" >&2
+    exit 1
+fi
+for want in init etc/inittab usr/local/bin/nats-server dev/console dev/null \
+    usr/local/lib/sqs-store/sqs-store-lib.sh usr/local/lib/sqs-store/sqs-store-prepare \
+    usr/local/lib/sqs-store/sqs-store-run usr/local/lib/sqs-store/sqs-store-stop; do
+    if ! grep -qx "$want" <<< "$archive_list"; then
+        echo "ERROR: $want missing from $INITRAMFS" >&2
+        exit 1
+    fi
+done
+for fs_tool in mkfs.ext4 e2fsck; do
+    if ! grep -qxE "(usr/)?sbin/$fs_tool" <<< "$archive_list"; then
+        echo "ERROR: $fs_tool missing from $INITRAMFS" >&2
+        exit 1
+    fi
+done
+# Module files may spell a name with dashes (tiny-power-button.ko). Normalise
+# once: piping into grep -q under pipefail fails when grep exits on a match.
+module_list=${archive_list//-/_}
+for want_ko in ext4 tiny_power_button; do
+    if ! grep -q "/$want_ko\.ko\$" <<< "$module_list"; then
+        echo "ERROR: $want_ko missing from $INITRAMFS" >&2
+        exit 1
+    fi
+done
+if grep -qE '(^|/)(lb-agent|haproxy|nginx)$' <<< "$archive_list"; then
+    echo "ERROR: LB data-plane binaries found in the SQS store image" >&2
+    exit 1
+fi
+echo "$tag content gate: PASS"
+
+echo ""
+echo "$tag staged artifacts:"
+ls -lh "$VMLINUZ" "$INITRAMFS"
+echo ""
+
+# --- Artifact size gate (50 MiB total) ---
+ARTIFACT_MAX_MiB="${MICROVM_ARTIFACT_MAX_MiB:-50}"
+total_bytes=$(( $(stat -c %s "$VMLINUZ") + $(stat -c %s "$INITRAMFS") ))
+max_bytes=$(( ARTIFACT_MAX_MiB * 1024 * 1024 ))
+total_mib=$(( (total_bytes + 1048575) / 1048576 ))
+if [ "$total_bytes" -gt "$max_bytes" ]; then
+    echo "ERROR: artifact size ${total_mib} MiB exceeds ${ARTIFACT_MAX_MiB} MiB gate" >&2
+    echo "       Set MICROVM_ARTIFACT_MAX_MiB=N to override during development." >&2
+    exit 1
+fi
+echo "$tag artifact size gate: PASS (${total_mib} MiB / ${ARTIFACT_MAX_MiB} MiB)"
+
+# --- Guest-RAM initramfs gate (uncompressed) ---
+# The kernel unpacks the whole initramfs into RAM before init, so the
+# UNCOMPRESSED size — not the gzip artifact — sets the guest memory floor.
+UNCOMPRESSED_MAX_MiB="${MICROVM_INITRAMFS_UNCOMPRESSED_MAX_MiB:-96}"
+uncompressed_bytes=$(zcat "$INITRAMFS" | wc -c)
+uncompressed_mib=$(( (uncompressed_bytes + 1048575) / 1048576 ))
+umax_bytes=$(( UNCOMPRESSED_MAX_MiB * 1024 * 1024 ))
+if [ "$uncompressed_bytes" -gt "$umax_bytes" ]; then
+    echo "ERROR: uncompressed initramfs ${uncompressed_mib} MiB exceeds ${UNCOMPRESSED_MAX_MiB} MiB guest-RAM gate" >&2
+    echo "       The kernel unpacks this into guest RAM before init; a larger" >&2
+    echo "       footprint panics before userspace. Slim the initramfs, or raise the" >&2
+    echo "       instance memory AND MICROVM_INITRAMFS_UNCOMPRESSED_MAX_MiB together." >&2
+    exit 1
+fi
+echo "$tag initramfs guest-RAM gate: PASS (${uncompressed_mib} MiB / ${UNCOMPRESSED_MAX_MiB} MiB uncompressed)"
+
+# --- Publish ---
+# Two renames are not atomic as a pair, so the old initramfs goes first and the
+# new one lands last. Stopped in between, the initramfs is absent: no mixed pair
+# exists and the grouped make target is missing, so make rebuilds.
+mkdir -p "$OUT_DIR"
+rm -f "$OUT_DIR/initramfs.cpio.gz"
+mv -f "$VMLINUZ" "$OUT_DIR/vmlinuz"
+mv -f "$INITRAMFS" "$OUT_DIR/initramfs.cpio.gz"
+echo "$tag published: $OUT_DIR/vmlinuz $OUT_DIR/initramfs.cpio.gz"
@@ -0,0 +1,196 @@
+#!/bin/sh
+# sqs-store-microvm-publish_test.sh — a staged SQS store microVM pair reaches
+# the output directory only when every gate passes, and a failed build cannot
+# leave a pair that make then takes as up to date.
+#
+# Artifacts are small fake archives listing the gated paths; no image is built.
+set -eu
+
+LIB_DIR=$(CDPATH='' cd -- "$(dirname -- "$0")" && pwd)
+PUBLISH="$LIB_DIR/sqs-store-microvm-publish.sh"
+TMP=$(mktemp -d)
+trap 'rm -rf "$TMP"' EXIT
+
+fails=0
+pass() { echo "  ok   - $1"; }
+fail() { echo "  FAIL - $1"; fails=$((fails + 1)); }
+
+echo "-- lib/sqs-store-microvm-publish.sh"
+
+for tool in cpio gzip make; do
+    if ! command -v "$tool" >/dev/null 2>&1; then
+        echo "  FAIL - required tool not found: $tool"
+        exit 1
+    fi
+done
+
+# stage <dir> [omit-path] — a fake staged pair; omit-path is left out of the archive.
+stage() {
+    dir=$1 omit=${2:-}
+    root="$TMP/root.$$.$(date +%s%N)"
+    for p in init etc/inittab usr/local/bin/nats-server sbin/mkfs.ext4 sbin/e2fsck \
+        usr/local/lib/sqs-store/sqs-store-lib.sh usr/local/lib/sqs-store/sqs-store-prepare \
+        usr/local/lib/sqs-store/sqs-store-run usr/local/lib/sqs-store/sqs-store-stop \
+        dev/console dev/null lib/modules/6.12-virt/kernel/fs/ext4/ext4.ko \
+        lib/modules/6.12-virt/kernel/drivers/acpi/tiny-power-button.ko; do
+        [ "$p" = "$omit" ] && continue
+        mkdir -p "$root/$(dirname "$p")"
+        echo "$p" > "$root/$p"
+    done
+    # pad=N appends N files after the gated ones, as a real image's listing has.
+    if [ "${pad:-0}" -gt 0 ]; then
+        mkdir -p "$root/zz"
+        (cd "$root/zz" && seq "$pad" | xargs touch)
+    fi
+    mkdir -p "$dir"
+    (cd "$root" && find . -mindepth 1 | sed 's|^\./||' | sort | cpio --quiet -o -H newc) | gzip > "$dir/initramfs.cpio.gz"
+    echo kernel > "$dir/vmlinuz"
+    rm -rf "$root"
+}
+
+# fingerprint <dir> — content and mtime of both outputs, or "absent".
+fingerprint() {
+    for f in "$1/vmlinuz" "$1/initramfs.cpio.gz"; do
+        if [ -e "$f" ]; then
+            printf '%s %s ' "$(cksum < "$f")" "$(stat -c %Y "$f")"
+        else
+            printf 'absent '
+        fi
+    done
+}
+
+# --- A valid pair is published and leaves the stage ---
+stage "$TMP/ok/stage"
+if bash "$PUBLISH" "$TMP/ok/stage" "$TMP/ok/out" >/dev/null 2>&1 \
+    && [ -s "$TMP/ok/out/vmlinuz" ] && [ -s "$TMP/ok/out/initramfs.cpio.gz" ] \
+    && [ ! -e "$TMP/ok/stage/vmlinuz" ] && [ ! -e "$TMP/ok/stage/initramfs.cpio.gz" ]; then
+    pass "valid pair is published"
+else
+    fail "valid pair is published"
+fi
+
+# --- A large listing with the modules early still passes the module gate ---
+pad=20000 stage "$TMP/large/stage"
+if bash "$PUBLISH" "$TMP/large/stage" "$TMP/large/out" >/dev/null 2>&1; then
+    pass "large archive with early modules is published"
+else
+    fail "large archive with early modules is published"
+fi
+
+# --- Each failing gate keeps the previous valid pair exactly ---
+for case in nats-server store-stop ext4 size; do
+    d="$TMP/keep-$case"
+    stage "$d/prev"
+    bash "$PUBLISH" "$d/prev" "$d/out" >/dev/null 2>&1
+    before=$(fingerprint "$d/out")
+    max=50
+    case $case in
+        nats-server) stage "$d/stage" usr/local/bin/nats-server ;;
+        store-stop) stage "$d/stage" usr/local/lib/sqs-store/sqs-store-stop ;;
+        ext4) stage "$d/stage" lib/modules/6.12-virt/kernel/fs/ext4/ext4.ko ;;
+        size) stage "$d/stage"; max=0 ;;
+    esac
+    if MICROVM_ARTIFACT_MAX_MiB=$max bash "$PUBLISH" "$d/stage" "$d/out" >/dev/null 2>&1; then
+        fail "$case gate failure is refused"
+    elif [ "$(fingerprint "$d/out")" = "$before" ]; then
+        pass "$case gate failure keeps the previous pair"
+    else
+        fail "$case gate failure keeps the previous pair"
+    fi
+done
+
+# --- With no previous pair, a failed gate publishes nothing ---
+stage "$TMP/none/stage" usr/local/bin/nats-server
+if bash "$PUBLISH" "$TMP/none/stage" "$TMP/none/out" >/dev/null 2>&1; then
+    fail "failed gate with no previous pair is refused"
+elif [ "$(fingerprint "$TMP/none/out")" = "absent absent " ]; then
+    pass "failed gate with no previous pair publishes nothing"
+else
+    fail "failed gate with no previous pair publishes nothing"
+fi
+
+# --- make cannot take a failed rebuild's pair as up to date ---
+# The same grouped-target shape as the Makefile; the builder stages a pair
+# missing nats-server once BAD exists, the way a failed real build would.
+m="$TMP/make"
+mkdir -p "$m/out"
+cat > "$m/build.sh" <<EOF
+#!/bin/sh
+set -eu
+rm -rf "$m/stage"
+. "$m/build.sh.lib"
+if [ -e "$m/BAD" ]; then
+    stage "$m/stage" usr/local/bin/nats-server
+else
+    stage "$m/stage"
+fi
+bash "$PUBLISH" "$m/stage" "$m/out"
+EOF
+{
+    echo "TMP=\"$TMP\""
+    sed -n '/^stage() {/,/^}/p' "$0"
+} > "$m/build.sh.lib"
+printf 'out/vmlinuz out/initramfs.cpio.gz &: input\n\tsh build.sh\n' > "$m/Makefile"
+echo v1 > "$m/input"
+if (cd "$m" && make -s out/vmlinuz >/dev/null 2>&1) && [ -s "$m/out/vmlinuz" ]; then
+    pass "make builds the first valid pair"
+else
+    fail "make builds the first valid pair"
+    echo "  $fails failure(s)"; exit 1
+fi
+good=$(fingerprint "$m/out")
+touch "$m/BAD"
+# Age the published pair so the input is strictly newer, without a future mtime.
+past="@$(( $(date +%s) - 10 ))"
+touch -d "$past" "$m/out/vmlinuz" "$m/out/initramfs.cpio.gz"
+touch "$m/input"
+good=$(fingerprint "$m/out")
+if (cd "$m" && make -s out/vmlinuz >/dev/null 2>&1); then
+    fail "make reports the failed rebuild"
+else
+    pass "make reports the failed rebuild"
+fi
+if (cd "$m" && make -q out/vmlinuz out/initramfs.cpio.gz); then
+    fail "after a failed rebuild make still sees the pair as stale"
+else
+    pass "after a failed rebuild make still sees the pair as stale"
+fi
+if [ "$(fingerprint "$m/out")" = "$good" ]; then
+    pass "failed rebuild leaves the previous valid pair"
+else
+    fail "failed rebuild leaves the previous valid pair"
+fi
+
+# --- A publish stopped after the kernel lands leaves no mixed pair ---
+# A stub mv lets the first rename through and fails the second: the point at
+# which an interrupt would leave a new kernel beside an old initramfs.
+i="$TMP/interrupt"
+mkdir -p "$i/bin"
+real_mv=$(command -v mv)
+{
+    echo '#!/bin/sh'
+    echo "if [ -e '$i/moved' ]; then exit 1; fi"
+    echo ": > '$i/moved'"
+    echo "exec '$real_mv' \"\$@\""
+} > "$i/bin/mv"
+chmod +x "$i/bin/mv"
+stage "$i/prev"
+bash "$PUBLISH" "$i/prev" "$i/out" >/dev/null 2>&1
+stage "$i/stage"
+echo new-kernel > "$i/stage/vmlinuz"
+if PATH="$i/bin:$PATH" bash "$PUBLISH" "$i/stage" "$i/out" >/dev/null 2>&1; then
+    fail "interrupted publish is reported"
+elif [ "$(cat "$i/out/vmlinuz")" = new-kernel ] && [ ! -e "$i/out/initramfs.cpio.gz" ]; then
+    pass "interrupted publish leaves no initramfs beside the new kernel"
+else
+    fail "interrupted publish leaves no initramfs beside the new kernel"
+fi
+printf 'out/vmlinuz out/initramfs.cpio.gz &: input\n\ttrue\n' > "$i/Makefile"
+touch -d "@$(( $(date +%s) - 10 ))" "$i/input"
+if (cd "$i" && make -q out/vmlinuz out/initramfs.cpio.gz); then
+    fail "make does not accept the interrupted publish"
+else
+    pass "make does not accept the interrupted publish"
+fi
+
+[ "$fails" -eq 0 ] || { echo "  $fails failure(s)"; exit 1; }
@@ -0,0 +1,18 @@
+#!/bin/sh
+# stage-sqs-store-microvm.sh <image-dir> <dist-root>: copy a built SQS store
+# microVM image into a release tree as <dist-root>/microvm-sqs-store. Without a
+# complete image it stages nothing and says so, as the load balancer image does.
+set -eu
+src=$1 dest=$2/microvm-sqs-store
+# An image staged by an earlier build must not ship with this one. Only the
+# two image files are removed; anything else in the directory is left alone.
+rm -f "$dest/vmlinuz" "$dest/initramfs.cpio.gz"
+if [ -s "$src/vmlinuz" ] && [ -s "$src/initramfs.cpio.gz" ]; then
+    echo "[distro] staging the SQS store microVM image into the tarball"
+    mkdir -p "$dest"
+    cp "$src/vmlinuz" "$src/initramfs.cpio.gz" "$dest/"
+else
+    [ ! -d "$dest" ] || rmdir "$dest" 2> /dev/null || true
+    echo "[distro] WARNING: SQS store microVM image missing; the tarball will not include it"
+    echo "[distro]          run 'make build-sqs-store-microvm-image' first"
+fi
@@ -190,7 +190,7 @@ run sudo systemctl reset-failed 'spinifex-*' 2>/dev/null || true
 # the wipe and before the verification, failing the reset on a file the script
 # had already deleted. update-nodes.sh carries the same wait.
 if ! $DRY_RUN; then
-    for unit in spinifex-predastore spinifex-viperblock spinifex-sqs-store spinifex-nats; do
+    for unit in spinifex-predastore spinifex-viperblock spinifex-nats; do
        if systemctl is-active --quiet "$unit"; then
            sudo systemctl stop "$unit" || true
        fi
@@ -273,6 +273,33 @@ if [ -n "$OVS_UNIT" ]; then
 fi
 run sudo rm -f /etc/openvswitch/system-id.conf

+# --- BEGIN ovs nss nicknames ---
+# ovs_nss_nicknames — the entries ovs-monitor-ipsec imported, read from
+# `certutil -L` on stdin: ovs_certkey_* is this node's certificate and key,
+# ovs_cert_* the CA and any peer certificates. Everything else is someone else's.
+ovs_nss_nicknames() {
+    awk '$1 ~ /^ovs_cert(key)?_/ {print $1}' | LC_ALL=C sort -u
+}
+# --- END ovs nss nicknames ---
+
+# A node rejoining another cluster must not keep trusting this one's CA. Only
+# the NSS entries go: other identities and /etc/ipsec.d/*.conf are left alone.
+if [ -x /usr/libexec/ipsec/pluto ] && command -v certutil >/dev/null 2>&1; then
+    log "removing Spinifex IPsec identities from libreswan's NSS database"
+    # A running ovs-monitor-ipsec imports them again from other_config.
+    run sudo systemctl stop openvswitch-ipsec.service 2>/dev/null || true
+    for nss in /var/lib/ipsec/nss /etc/ipsec.d; do
+        sudo test -e "$nss/cert9.db" || continue
+        for nick in $(sudo certutil -L -d "sql:$nss" 2>/dev/null | ovs_nss_nicknames); do
+            log "  deleting $nick from $nss"
+            delete=-D
+            case "$nick" in ovs_certkey_*) delete=-F ;; esac
+            run sudo certutil "$delete" -d "sql:$nss" -n "$nick" ||
+                log "  NOTE: could not delete $nick from $nss; remove it with certutil by hand"
+        done
+    done
+fi
+
 # Delete the OVN DBs outright — a caller re-running setup-ovn.sh gets fresh
 # empty ones. This is what clears stale chassis rows and port bindings, which
 # otherwise accumulate across resets and wedge ovn-controller in a commit loop.
@@ -401,10 +428,6 @@ if ! $KEEP_DATA && [ -d "$DATA_DIR/nats/jetstream" ]; then
    log "removing the JetStream store"
    run sudo rm -rf "$DATA_DIR/nats/jetstream"
 fi
-if ! $KEEP_DATA && [ -d "$DATA_DIR/sqs-store/jetstream" ]; then
-    log "removing the SQS store JetStream data"
-    run sudo rm -rf "$DATA_DIR/sqs-store/jetstream"
-fi

 # Directories may survive — a mountpoint cannot be removed, and neither can the
 # path leading down to one. Files may not: a surviving file is state that would
@@ -1,7 +1,8 @@
 #!/bin/sh
 # node-reset_test.sh — tests the QEMU process match node-reset.sh uses to stop
 # guests before it deletes Viperblock state, against real processes, and the
-# instance and mounted-volume counts it and uninstall-spx.sh report before deleting.
+# instance and mounted-volume counts it and uninstall-spx.sh report before deleting,
+# and the libreswan NSS entries it deletes.
 #
 # Each fake guest is a copy of sleep named like a QEMU binary, so the kernel
 # gives it the same comm, including the 15-character truncation.
@@ -129,6 +130,32 @@ mkdir -p "$empty"
 expect "counts no instances without a state file" "$(count count_instances "$empty")" 0
 expect "counts no mounted volumes without a viperblock directory" "$(count count_mounted_volumes "$empty")" 0

+echo "-- node-reset.sh libreswan NSS nicknames"
+
+sed -n '/^# --- BEGIN ovs nss nicknames ---$/,/^# --- END ovs nss nicknames ---$/p' \
+    "$SCRIPT_DIR/node-reset.sh" > "$TMP/nss.sh"
+# `certutil -L` as it prints a database holding an operator's own identity and
+# CA beside what ovs-monitor-ipsec imported, one peer listed twice.
+cat > "$TMP/certutil-L" <<'EOF'
+
+Certificate Nickname                                         Trust Attributes
+                                                             SSL,S/MIME,JAR/XPI
+
+ovs_certkey_node1                                            u,u,u
+ovs_cert_cacert                                              CT,,
+ovs_cert_node2                                               P,P,P
+ovs_cert_node2                                               P,P,P
+site-vpn                                                     u,u,u
+corp-ca                                                      CT,C,C
+ovs_certificate_store                                        ,,
+EOF
+if [ ! -s "$TMP/nss.sh" ]; then
+    fail "no ovs nss nicknames block in node-reset.sh"
+else
+    got=$(bash -c '. "$1"; ovs_nss_nicknames < "$2"' _ "$TMP/nss.sh" "$TMP/certutil-L" | tr '\n' ' ')
+    expect "selects only ovs-monitor-ipsec's entries, once each" "$got" "ovs_cert_cacert ovs_cert_node2 ovs_certkey_node1 "
+fi
+
 if [ "$fails" -gt 0 ]; then
    echo "$fails failure(s)"
    exit 1
@@ -172,6 +172,8 @@ OVN_DBDIR="${OVN_DBDIR:-/var/lib/ovn}"
 # Seconds to wait for the database servers to exit, then for a SIGKILL to land.
 OVN_DB_STOP_WAIT="${OVN_DB_STOP_WAIT:-30}"
 OVN_DB_KILL_WAIT="${OVN_DB_KILL_WAIT:-10}"
+# Seconds to wait for a restarted clustered NB or SB member to rejoin its cluster.
+OVN_RAFT_REJOIN_WAIT="${OVN_RAFT_REJOIN_WAIT:-60}"
 # Records that --management set this node up. Nodes set up before it existed
 # carry only the legacy marker, OVN_CTL_OPTS in /etc/default/ovn-central.
 OVN_ROLE_FILE=/etc/spinifex/ovn-role
@@ -757,6 +759,42 @@ stop_ovn_databases() {
    exit 1
 }

+# db_clients — sets NBCTL/SBCTL. A clustered DB has no unix socket to fall back
+# on, and a follower refuses to answer at all unless leader-only is turned off;
+# writes through a follower are forwarded to the leader.
+db_clients() {
+    NBCTL=(ovn-nbctl)
+    SBCTL=(ovn-sbctl)
+    if [ -n "$DB_CLUSTER_LOCAL_ADDR" ]; then
+        NBCTL=(ovn-nbctl --db="tcp:$DB_CLUSTER_LOCAL_ADDR:6641" --no-leader-only)
+        SBCTL=(ovn-sbctl --db="tcp:$DB_CLUSTER_LOCAL_ADDR:6642" --no-leader-only)
+    fi
+}
+
+# set_client_connections — the loopback client listeners. A rerun's restart can
+# hand the leadership away, so this goes through NBCTL/SBCTL, not the socket.
+set_client_connections() {
+    sudo "${NBCTL[@]}" set-connection ptcp:6641:127.0.0.1
+    sudo "${SBCTL[@]}" set-connection ptcp:6642:127.0.0.1
+}
+
+# wait_for_raft_member <ctl> <schema> — waits until this server is a cluster
+# member that knows its leader, which a restarted member is once it rejoins.
+# Returns 1 on timeout.
+wait_for_raft_member() {
+    local ctl=$1 schema=$2 status i
+    for i in $(seq 1 "$OVN_RAFT_REJOIN_WAIT"); do
+        status=$(sudo ovs-appctl -t "$ctl" cluster/status "$schema" 2>/dev/null || true)
+        if grep -qx 'Status: cluster member' <<<"$status" &&
+            grep -q '^Leader: ' <<<"$status" && ! grep -qx 'Leader: unknown' <<<"$status"; then
+            return 0
+        fi
+        echo "  Waiting for $schema to rejoin its RAFT cluster... ($i/$OVN_RAFT_REJOIN_WAIT)"
+        sleep 1
+    done
+    return 1
+}
+
 # --- Step 2: Enable services ---
 # ovn-ctl consults the RAFT flags only when it creates a database, so any
 # existing DB file is served as it stands and the cluster flags are ignored.
@@ -996,16 +1034,9 @@ EOF
        fi
    fi

-    # A clustered DB has no unix socket to fall back on, and a follower refuses
-    # to answer at all unless leader-only is turned off. Without both of these
-    # the waits below can never succeed on a clustered node: they burn their
-    # full timeout, gate nothing, and the run still continues.
-    NBCTL=(ovn-nbctl)
-    SBCTL=(ovn-sbctl)
-    if [ -n "$DB_CLUSTER_LOCAL_ADDR" ]; then
-        NBCTL=(ovn-nbctl --db="tcp:$DB_CLUSTER_LOCAL_ADDR:6641" --no-leader-only)
-        SBCTL=(ovn-sbctl --db="tcp:$DB_CLUSTER_LOCAL_ADDR:6642" --no-leader-only)
-    fi
+    # Without these the waits below can never succeed on a clustered node:
+    # they burn their full timeout, gate nothing, and the run still continues.
+    db_clients

    # Wait for OVN NB DB socket to become available
    for i in $(seq 1 15); do
@@ -1067,29 +1098,34 @@ EOF
    # with no join step to absorb the window — attaching before the leader is
    # elected (and before northd writes the gateway logical flows) leaves it with
    # a partial datapath that never reconverges. Gate on SB reachable and, when
-    # clustered, an elected leader.
+    # clustered, both members back in their clusters under an elected leader.
    for i in $(seq 1 30); do
-        if sudo "${SBCTL[@]}" --timeout=2 show >/dev/null 2>&1; then
-            if [ -z "$DB_CLUSTER_LOCAL_ADDR" ]; then
-                break
-            fi
-            SB_LEADER=$(sudo ovs-appctl -t /var/run/ovn/ovnsb_db.ctl \
-                cluster/status OVN_Southbound 2>/dev/null \
-                | awk '/^Leader:/{print $2; exit}')
-            if [ -n "$SB_LEADER" ] && [ "$SB_LEADER" != "unknown" ]; then
-                break
-            fi
-        fi
-        echo "  Waiting for OVN SB DB leader... ($i/30)"
+        sudo "${SBCTL[@]}" --timeout=2 show >/dev/null 2>&1 && break
+        echo "  Waiting for OVN SB DB... ($i/30)"
        sleep 1
    done

+    # The restart above took both of this node's members down at once. Returning
+    # before they rejoin lets a rerun on the next database node take a second
+    # member of each cluster down, and one of three left is below quorum.
+    if [ -n "$DB_CLUSTER_LOCAL_ADDR" ]; then
+        for member in "/var/run/ovn/ovnnb_db.ctl OVN_Northbound" "/var/run/ovn/ovnsb_db.ctl OVN_Southbound"; do
+            # shellcheck disable=SC2086 # two words: the ctl socket and the schema
+            if ! wait_for_raft_member $member; then
+                echo "ERROR: this node's ${member#* } server is not back in its RAFT cluster after ${OVN_RAFT_REJOIN_WAIT}s." >&2
+                echo "       Restarting OVN central on another database node now would leave the cluster" >&2
+                echo "       without quorum. Check: sudo ovs-appctl -t ${member% *} cluster/status ${member#* }" >&2
+                exit 1
+            fi
+        done
+        echo "  RAFT: NB and SB members rejoined, leader elected"
+    fi
+
    # set-connection replicates through RAFT, so it only runs on the create
    # node. Scoped to loopback, not a wildcard ptcp:6641, because 127.0.0.1
    # is valid on every node; the differing LAN address is set above instead.
    if [ -z "$DB_CLUSTER_REMOTE_ADDR" ]; then
-        sudo ovn-nbctl set-connection ptcp:6641:127.0.0.1
-        sudo ovn-sbctl set-connection ptcp:6642:127.0.0.1
+        set_client_connections
        if [ -n "$LAN_ADDR" ]; then
            echo "  OVN NB DB listening on tcp:127.0.0.1:6641, tcp:$LAN_ADDR:6641"
            echo "  OVN SB DB listening on tcp:127.0.0.1:6642, tcp:$LAN_ADDR:6642"
@@ -1958,16 +1994,21 @@ echo "Step 10: Enabling OVN auto-start on boot..."
 sudo systemctl enable "$OVS_UNIT" 2>/dev/null || true
 sudo systemctl enable ovn-controller 2>/dev/null || true
 printf '  %-21s enabled on boot\n' "${OVS_UNIT%.service}:" ovn-controller:
-if distro_has_libreswan; then
-    # OVN IPsec does not run on libreswan yet, so it stays off here. Starting
-    # openvswitch-ipsec would also start pluto, which listens on UDP 500/4500.
+if distro_has_libreswan && [ "$(systemctl is-active openvswitch-ipsec.service 2>/dev/null)" = active ]; then
+    # The daemon turned IPsec on; stopping it here would drop live tunnels.
+    printf '  %-21s %s\n' openvswitch-ipsec: "active: IPsec is on, left to the daemon" ipsec: "active: IPsec is on, left to the daemon"
+elif distro_has_libreswan; then
+    # Off until the daemon asks: pluto listens on UDP 500/4500 once started.
+    # Enabled under the mask, so the daemon's unmask also brings both back on boot.
    IPSEC_UNITS=(ipsec.service)
    distro_unit_exists openvswitch-ipsec.service && IPSEC_UNITS+=(openvswitch-ipsec.service)
-    if ! sudo systemctl mask --now "${IPSEC_UNITS[@]}" >/dev/null; then
-        echo "ERROR: could not stop and mask ${IPSEC_UNITS[*]}; IPsec must stay off on libreswan" >&2
+    if ! sudo systemctl unmask "${IPSEC_UNITS[@]}" >/dev/null 2>&1 ||
+        ! sudo systemctl enable "${IPSEC_UNITS[@]}" >/dev/null 2>&1 ||
+        ! sudo systemctl mask --now "${IPSEC_UNITS[@]}" >/dev/null; then
+        echo "ERROR: could not enable, stop and mask ${IPSEC_UNITS[*]}; IPsec must stay off until the daemon turns it on" >&2
        exit 1
    fi
-    printf '  %-21s %s\n' openvswitch-ipsec: "masked: IPsec stays off on libreswan" ipsec: "masked: IPsec stays off on libreswan"
+    printf '  %-21s %s\n' openvswitch-ipsec: "masked: IPsec is off, left to the daemon" ipsec: "masked: IPsec is off, left to the daemon"
 elif [ "$(systemctl is-enabled openvswitch-ipsec.service 2>/dev/null)" = masked ]; then
    # The daemon masks it while IPsec is off and unmasks it to turn IPsec on.
    printf '  %-21s %s\n' openvswitch-ipsec: "masked: IPsec is off, left to the daemon"
@@ -454,6 +454,47 @@ case "$RC:$OUT" in
    *) fail "no recreate-db: standalone DB rc=$RC out=$OUT" ;;
 esac

+# --------------------------------------------------------------------------
+# A rerun returns only once both of this node's RAFT members have rejoined.
+# --------------------------------------------------------------------------
+# A rerun restarts the node's NB and SB members together, so returning early
+# lets the next database node's rerun take a second member of each down.
+awk '$0 == "wait_for_raft_member() {" { on = 1 } on { print } on && $0 == "}" { exit }' "$SETUP" >"$TMP/raft_fn.sh"
+mkdir -p "$TMP/raftbin"
+printf '#!/bin/sh\nexit 0\n' >"$TMP/raftbin/sleep"
+printf '#!/bin/sh\n[ -f "%s" ] || exit 1\ncat "%s"\n' "$TMP/cluster-status" "$TMP/cluster-status" >"$TMP/raftbin/ovs-appctl"
+chmod +x "$TMP/raftbin/sleep" "$TMP/raftbin/ovs-appctl"
+
+# raft_member <Status:> <Leader:> — runs wait_for_raft_member against a
+# cluster/status in that state, or none when Status is "absent"; sets RC.
+raft_member() {
+    rm -f "$TMP/cluster-status"
+    if [ "$1" != absent ]; then
+        printf 'a1b2\nName: OVN_Northbound\nStatus: %s\nRole: follower\nTerm: 4\nLeader: %s\nVote: c3d4\n\nServers:\n    a1b2 (a1b2 at tcp:10.0.0.1:6643) (self)\n    c3d4 (c3d4 at tcp:10.0.0.2:6643)\n' \
+            "$1" "$2" >"$TMP/cluster-status"
+    fi
+    RC=0
+    PATH="$TMP/raftbin:$PATH" bash -c 'sudo() { "$@"; }; OVN_RAFT_REJOIN_WAIT=2; . "$1"
+        wait_for_raft_member /var/run/ovn/ovnnb_db.ctl OVN_Northbound' _ "$TMP/raft_fn.sh" >/dev/null 2>&1 || RC=$?
+}
+
+if [ ! -s "$TMP/raft_fn.sh" ]; then
+    fail "raft: no wait_for_raft_member in setup-ovn.sh"
+else
+    raft_member "cluster member" c3d4
+    if [ "$RC" = 0 ]; then pass "raft: a member following a known leader has rejoined"; else fail "raft: follower with leader rc=$RC"; fi
+    raft_member "cluster member" self
+    if [ "$RC" = 0 ]; then pass "raft: a member that is the leader has rejoined"; else fail "raft: leader rc=$RC"; fi
+    raft_member "cluster member" unknown
+    if [ "$RC" != 0 ]; then pass "raft: a member that has not found its leader has not rejoined"; else fail "raft: unknown leader accepted"; fi
+    raft_member "disconnected from the cluster (election timeout)" c3d4
+    if [ "$RC" != 0 ]; then pass "raft: a disconnected member that remembers a leader has not rejoined"; else fail "raft: disconnected member accepted"; fi
+    raft_member "joining cluster" c3d4
+    if [ "$RC" != 0 ]; then pass "raft: a member still joining has not rejoined"; else fail "raft: joining member accepted"; fi
+    raft_member absent ""
+    if [ "$RC" != 0 ]; then pass "raft: a server that does not answer has not rejoined"; else fail "raft: unanswered cluster/status accepted"; fi
+fi
+
 rm -rf "$TMP/dbdir"
 mkdir -p "$TMP/dbdir"
 ensure_db true
@@ -478,6 +519,33 @@ else
 fi
 printf '#!/bin/sh\nexit 1\n' >"$TMP/bin/pgrep"

+# A rerun restarts this node's members, which can hand the leadership to a
+# peer, so the create node's set-connection must still land as a follower.
+for fn in db_clients set_client_connections; do
+    awk -v f="$fn() {" '$0 == f { on = 1 } on { print } on && $0 == "}" { exit }' "$SETUP" >>"$TMP/conn_fn.sh"
+done
+mkdir -p "$TMP/followerbin"
+for ctl in ovn-nbctl ovn-sbctl; do
+    cat >"$TMP/followerbin/$ctl" <<EOF
+#!/bin/sh
+case " \$* " in
+    *" --no-leader-only "*) echo "$ctl \$*" >>"$TMP/conn.log" ;;
+    *) echo "$ctl: unix:/var/run/ovn/db.sock: database connection failed ()" >&2; exit 1 ;;
+esac
+EOF
+    chmod +x "$TMP/followerbin/$ctl"
+done
+: >"$TMP/conn.log"
+RC=0
+PATH="$TMP/followerbin:$PATH" bash -c 'set -e; sudo() { "$@"; }; DB_CLUSTER_LOCAL_ADDR=10.0.0.1; . "$1"
+    db_clients; set_client_connections' _ "$TMP/conn_fn.sh" >/dev/null 2>&1 || RC=$?
+if [ "$RC" = 0 ] && [ "$(cat "$TMP/conn.log")" = "ovn-nbctl --db=tcp:10.0.0.1:6641 --no-leader-only set-connection ptcp:6641:127.0.0.1
+ovn-sbctl --db=tcp:10.0.0.1:6642 --no-leader-only set-connection ptcp:6642:127.0.0.1" ]; then
+    pass "connections: the create node sets the client listeners while a follower"
+else
+    fail "connections: as a follower rc=$RC calls [$(cat "$TMP/conn.log")]"
+fi
+
 if [ "$fails" -ne 0 ]; then
    echo "  setup-ovn.sh: $fails failure(s)"
    exit 1
@@ -283,7 +283,6 @@ create_service_users() {
    # Create per-service users with correct home directories
    declare -A SERVICE_HOMES=(
        [nats]="/var/lib/spinifex/nats"
-        [sqsstore]="/var/lib/spinifex/sqs-store"
        [gw]="/var/lib/spinifex/awsgw"
        [daemon]="/var/lib/spinifex/spinifex"
        [storage]="/var/lib/spinifex/predastore"
@@ -292,7 +291,7 @@ create_service_users() {
        [vpcd]="/var/lib/spinifex"
        [ui]="/var/lib/spinifex"
    )
-    for svc in nats sqsstore gw daemon storage northstar viperblock vpcd ui; do
+    for svc in nats gw daemon storage northstar viperblock vpcd ui; do
        local user="spinifex-${svc}"
        if ! id "$user" > /dev/null 2>&1; then
            $SUDO useradd --system --no-create-home \
@@ -324,7 +323,7 @@ create_service_users() {
    # Daemon consumes viperblock's NBD socket — join the producer-typed group.
    $SUDO usermod -aG spinifex-viperblock spinifex-daemon

-    info "Service users created (spinifex-{nats,sqsstore,gw,daemon,storage,northstar,viperblock,vpcd,ui})"
+    info "Service users created (spinifex-{nats,gw,daemon,storage,northstar,viperblock,vpcd,ui})"
 }

 # --- Install scoped sudoers rules ---
@@ -390,10 +389,68 @@ install_ipsec_state_helper() {
 set -eu

 if [ "$#" -ne 1 ]; then
-    echo "usage: $0 <on|off>" >&2
+    echo "usage: $0 <on|off|guard|unguard>" >&2
    exit 2
 fi

+# libreswan drops a tunnel's xfrm policies while IKE fails, and pluto flushes
+# the whole xfrm table when it starts or stops, so Geneve from the encap address
+# is dropped in nft unless the route carries an SA.
+GUARD_TABLE="inet spinifex_ipsec_guard"
+
+unguard() {
+    printf 'table %s\ndelete table %s\n' "$GUARD_TABLE" "$GUARD_TABLE" | nft -f -
+}
+
+want=""
+want_src() {
+    case "$1" in
+        *.*.*.*) case "$1" in *[!0-9.]*) src=0.0.0.0/0 ;; *) src="$1/32" ;; esac ;;
+        *) src=0.0.0.0/0 ;;
+    esac
+    case " $want " in *" $src "*) ;; *) want="$want $src" ;; esac
+}
+
+# One source per local address of a tunnel with a remote_name, which
+# ovn-controller sets only while NB_Global.ipsec is on. A tunnel without a
+# usable local_ip is guarded from every source rather than left open.
+guard() {
+    ifaces=$(ovs-vsctl --timeout=5 --bare --columns=name find Interface type=geneve)
+    for iface in $ifaces; do
+        peer=$(ovs-vsctl --timeout=5 --if-exists get Interface "$iface" options:remote_name)
+        [ -n "$peer" ] || continue
+        want_src "$(ovs-vsctl --timeout=5 --if-exists get Interface "$iface" options:local_ip | tr -d '"')"
+    done
+    # At boot pluto is up before ovn-controller builds the tunnels, and its
+    # callers guard only while encryption is wanted, so guard the encap address
+    # until the tunnels say otherwise.
+    if [ -z "$ifaces" ]; then
+        encap=$(ovs-vsctl --timeout=5 --if-exists get Open_vSwitch . external_ids:ovn-encap-ip | tr -d '"')
+        for ip in $(echo "$encap" | tr ',' ' '); do want_src "$ip"; done
+        [ -n "$want" ] || want_src ""
+    fi
+    if [ -z "$want" ]; then
+        unguard
+        return
+    fi
+    # One transaction, so the old set is never gone before the new one is in.
+    nft -f - << NFT
+table $GUARD_TABLE
+delete table $GUARD_TABLE
+table $GUARD_TABLE {
+    set srcs {
+        type ipv4_addr
+        flags interval
+        elements = { $(echo $want | tr ' ' ',') }
+    }
+    chain out {
+        type filter hook output priority 0; policy accept;
+        ip saddr @srcs udp dport 6081 rt ipsec missing drop
+    }
+}
+NFT
+}
+
 # mask/unmask rather than disable/enable. The daemon's unit sets
 # ProtectSystem=full, so /etc is read-only in the namespace its children inherit,
 # and disabling a unit that also ships a SysV script shells out to update-rc.d
@@ -401,12 +458,24 @@ fi

 # ovs-monitor-ipsec execs the strongSwan starter itself rather than going through
 # this unit, so leaving it enabled only contends for UDP 500/4500. Off in both
-# states. Absent on a host without strongswan, which is not an error.
-systemctl mask --now strongswan-starter.service >/dev/null 2>&1 || true
+# states. Masking a unit no package ships leaves a stray /dev/null link, so a
+# host without strongSwan gets none, and loses one an older helper left.
+if [ -e /usr/lib/systemd/system/strongswan-starter.service ] || [ -e /lib/systemd/system/strongswan-starter.service ]; then
+    systemctl mask --now strongswan-starter.service >/dev/null 2>&1 || true
+elif [ "$(systemctl is-enabled strongswan-starter.service 2>/dev/null)" = masked ]; then
+    systemctl unmask strongswan-starter.service >/dev/null 2>&1 || true
+fi

 case "$1" in
    on)
-        systemctl unmask openvswitch-ipsec.service
+        if [ -x /usr/libexec/ipsec/pluto ]; then
+            # ovs-monitor-ipsec drives pluto over its control socket but does
+            # not start it, so pluto comes up first under its own unit.
+            systemctl unmask ipsec.service openvswitch-ipsec.service
+            systemctl start ipsec.service
+        else
+            systemctl unmask openvswitch-ipsec.service
+        fi
        exec systemctl start openvswitch-ipsec.service
        ;;
    off)
@@ -415,8 +484,15 @@ case "$1" in
        # On Debian that name is strongswan-starter's alias, masked above.
        if [ -x /usr/libexec/ipsec/pluto ]; then
            systemctl mask --now ipsec.service
+            unguard
        fi
        ;;
+    guard)
+        guard
+        ;;
+    unguard)
+        unguard
+        ;;
    *)
        echo "state not permitted: $1" >&2
        exit 2
@@ -624,7 +700,7 @@ table inet spinifex_filter {
        # inventory and running instances to anyone who asks. 8660 is
        # predastore's unauthenticated /healthz and /readyz, which name the
        # peers a gate cannot reach, so it belongs here for the same reason.
-        ip saddr $spinifex_peers tcp dport { 4222, 4223, 4248, 4249, 4432, 5300, 6641, 6642, 6643, 6644, 8660 } accept
+        ip saddr $spinifex_peers tcp dport { 4222, 4248, 4432, 5300, 6641, 6642, 6643, 6644, 8660 } accept
        ip saddr $spinifex_peers udp dport { 5300, 6660, 7660 } accept

        # Encap plane, peer-scoped: Geneve, IKE, NAT-T and ESP. Geneve is a
@@ -1466,6 +1542,41 @@ install_files() {
        $SUDO cp "$EXTRACT_DIR/microvm/"* /usr/share/spinifex/microvm/
        info "  /usr/share/spinifex/microvm/*"
    fi
+    install_sqs_store_image
+}
+
+# The SQS store image ships beside the load balancer's, not inside it: an older
+# installer copies microvm/* without -r and would stop at a subdirectory.
+install_sqs_store_image() {
+    local src="$EXTRACT_DIR/microvm-sqs-store" f
+    local dest="${SPINIFEX_MICROVM_DIR:-/usr/share/spinifex/microvm}/sqs-store"
+    # A previously installed image belongs to an older release, and this
+    # release's daemon must not boot it. Only the two image files are removed.
+    if [ ! -d "$src" ]; then
+        remove_sqs_store_image "$dest"
+        warn "This release has no SQS store microVM image; SQS store launches will be refused"
+        return 0
+    fi
+    for f in vmlinuz initramfs.cpio.gz; do
+        if [ ! -s "$src/$f" ]; then
+            remove_sqs_store_image "$dest"
+            fatal "The SQS store microVM image in this release is incomplete: $f is missing"
+        fi
+    done
+    # The initramfs goes last, so an interrupted install leaves no pair the
+    # daemon would boot: it refuses a launch while either file is missing.
+    $SUDO install -d "$dest"
+    $SUDO rm -f "$dest/initramfs.cpio.gz"
+    $SUDO install -m 0644 "$src/vmlinuz" "$dest/vmlinuz"
+    $SUDO install -m 0644 "$src/initramfs.cpio.gz" "$dest/initramfs.cpio.gz"
+    info "  $dest/vmlinuz, $dest/initramfs.cpio.gz"
+}
+
+# Running store VMs keep their already-loaded kernel; the files are read at boot.
+remove_sqs_store_image() {
+    [ -e "$1/initramfs.cpio.gz" ] || [ -e "$1/vmlinuz" ] || return 0
+    $SUDO rm -f "$1/initramfs.cpio.gz" "$1/vmlinuz"
+    warn "Removed the SQS store microVM image left by an earlier install from $1"
 }

 # --- Create directories ---
@@ -1521,12 +1632,6 @@ TMPEOF
    $SUDO chown "spinifex-nats:$SPINIFEX_GROUP" /etc/spinifex/nats
    $SUDO chmod 0750 /etc/spinifex/nats

-    # The SQS store holds its own token: 0700 keeps every other spinifex-*
-    # service in the shared group out.
-    $SUDO mkdir -p /etc/spinifex/sqs-store
-    $SUDO chown "spinifex-sqsstore:$SPINIFEX_GROUP" /etc/spinifex/sqs-store
-    $SUDO chmod 0700 /etc/spinifex/sqs-store
-
    $SUDO mkdir -p /etc/spinifex/predastore
    $SUDO chown "spinifex-storage:$SPINIFEX_GROUP" /etc/spinifex/predastore
    $SUDO chmod 0750 /etc/spinifex/predastore
@@ -1554,10 +1659,6 @@ TMPEOF
    $SUDO chown "spinifex-nats:$SPINIFEX_GROUP" /var/lib/spinifex/nats
    $SUDO chmod 0700 /var/lib/spinifex/nats

-    $SUDO mkdir -p /var/lib/spinifex/sqs-store
-    $SUDO chown "spinifex-sqsstore:$SPINIFEX_GROUP" /var/lib/spinifex/sqs-store
-    $SUDO chmod 0700 /var/lib/spinifex/sqs-store
-
    $SUDO mkdir -p /var/lib/spinifex/spinifex
    $SUDO chown "spinifex-daemon:$SPINIFEX_GROUP" /var/lib/spinifex/spinifex
    $SUDO chmod 0700 /var/lib/spinifex/spinifex
@@ -1639,7 +1740,6 @@ fix_file_ownership() {
    # Per-service data dirs — recursive chown so existing files are accessible
    for entry in \
        nats:spinifex-nats \
-        sqs-store:spinifex-sqsstore \
        predastore:spinifex-storage \
        northstar:spinifex-northstar \
        spinifex:spinifex-daemon \
@@ -1658,11 +1758,6 @@ fix_file_ownership() {
        $SUDO chown -R "spinifex-nats:$SPINIFEX_GROUP" /etc/spinifex/nats \
            || fatal "Failed to set ownership on /etc/spinifex/nats"
    fi
-    if [ -d /etc/spinifex/sqs-store ]; then
-        $SUDO chown -R "spinifex-sqsstore:$SPINIFEX_GROUP" /etc/spinifex/sqs-store \
-            || fatal "Failed to set ownership on /etc/spinifex/sqs-store"
-        $SUDO chmod 0700 /etc/spinifex/sqs-store
-    fi
    if [ -d /etc/spinifex/predastore ]; then
        $SUDO chown -R "spinifex-storage:$SPINIFEX_GROUP" /etc/spinifex/predastore \
            || fatal "Failed to set ownership on /etc/spinifex/predastore"
@@ -1769,6 +1864,18 @@ CPUWeight=300
 EOF
    info "  /etc/systemd/system/system.slice.d/spinifex-reserve.conf"

+    # The plaintext Geneve guard is not persisted, so it goes in when pluto
+    # starts at boot, before ovn-controller builds the tunnels. `-`: the guard
+    # must never fail the unit, and the daemon re-asserts it each pass.
+    if [ -x /usr/libexec/ipsec/pluto ]; then
+        $SUDO mkdir -p /etc/systemd/system/ipsec.service.d
+        $SUDO tee /etc/systemd/system/ipsec.service.d/spinifex-guard.conf > /dev/null << EOF
+[Service]
+ExecStartPost=-${IPSEC_STATE_HELPER} guard
+EOF
+        info "  /etc/systemd/system/ipsec.service.d/spinifex-guard.conf"
+    fi
+
    # daemon-reload / enable require a running systemd — skip inside the ISO
    # chroot. Unit files are still dropped into place; firstboot enables them.
    if [ "${ISO_BUILD:-0}" = "1" ]; then
@@ -0,0 +1,190 @@
+#!/bin/sh
+# setup_sqs_store_image_test.sh — tests for shipping the SQS store microVM
+# image: staging it into a release tree and installing it with setup.sh.
+#
+# setup.sh is sourced with INSTALL_SPINIFEX_LIB_ONLY=1 so main never runs, and
+# installs into a temporary SPINIFEX_MICROVM_DIR with SUDO cleared.
+set -eu
+
+SCRIPT_DIR=$(CDPATH='' cd -- "$(dirname -- "$0")" && pwd)
+SETUP="$SCRIPT_DIR/setup.sh"
+STAGE="$SCRIPT_DIR/lib/stage-sqs-store-microvm.sh"
+TMP=$(mktemp -d)
+trap 'rm -rf "$TMP"' EXIT
+
+fails=0
+pass() { echo "  ok   - $1"; }
+fail() { echo "  FAIL - $1"; fails=$((fails + 1)); }
+
+echo "-- SQS store microVM image release and install"
+
+# image <dir> <tag>: a fake built image whose bytes name its tag.
+image() {
+    mkdir -p "$1"
+    printf 'kernel-%s\n' "$2" > "$1/vmlinuz"
+    printf 'initramfs-%s\n' "$2" > "$1/initramfs.cpio.gz"
+}
+
+# release <dir>: a release tree with the load balancer image, as distro-amd64 builds it.
+release() {
+    rm -rf "$1"
+    mkdir -p "$1/microvm"
+    printf 'lb-kernel\n' > "$1/microvm/vmlinuz"
+    printf 'lb-initramfs\n' > "$1/microvm/initramfs.cpio.gz"
+    printf 'qboot\n' > "$1/microvm/qboot.rom"
+}
+
+# install_image <extract-dir> <microvm-dir> [path-prefix]: runs
+# install_sqs_store_image and sets RC and OUT.
+install_image() {
+    RC=0
+    OUT=$(PATH="${3:-}$PATH" EXTRACT_DIR="$1" SPINIFEX_MICROVM_DIR="$2" INSTALL_SPINIFEX_LIB_ONLY=1 \
+        bash -c '. "$1"; SUDO=""; install_sqs_store_image' _ "$SETUP" 2>&1) || RC=$?
+}
+
+same() { cmp -s "$1" "$2"; }
+
+# --- staging ---
+image "$TMP/built" v1
+release "$TMP/rel"
+sh "$STAGE" "$TMP/built" "$TMP/rel" > "$TMP/stage.out"
+if same "$TMP/built/vmlinuz" "$TMP/rel/microvm-sqs-store/vmlinuz" \
+    && same "$TMP/built/initramfs.cpio.gz" "$TMP/rel/microvm-sqs-store/initramfs.cpio.gz"; then
+    pass "stages the built image beside the load balancer's"
+else
+    fail "staged image differs from the build"
+fi
+if [ ! -e "$TMP/rel/microvm/sqs-store" ] && [ "$(ls "$TMP/rel/microvm" | wc -l)" -eq 3 ]; then
+    pass "leaves the load balancer's microvm directory as it was"
+else
+    fail "load balancer directory changed: $(ls "$TMP/rel/microvm")"
+fi
+
+release "$TMP/rel-none"
+mkdir -p "$TMP/half"; printf 'k\n' > "$TMP/half/vmlinuz"
+OUT=$(sh "$STAGE" "$TMP/half" "$TMP/rel-none")
+if [ ! -e "$TMP/rel-none/microvm-sqs-store" ] && echo "$OUT" | grep -q 'WARNING: SQS store microVM image missing'; then
+    pass "stages nothing from an incomplete build, and says so"
+else
+    fail "incomplete build staged or unannounced: $OUT"
+fi
+
+# --- an older installer meets a release that carries the store image ---
+# It copies microvm/* without -r; the sibling directory must not reach it.
+mkdir -p "$TMP/old-dest"
+if cp "$TMP/rel/microvm/"* "$TMP/old-dest/" 2> "$TMP/old.err" && [ ! -e "$TMP/old-dest/sqs-store" ]; then
+    pass "an older installer's copy still succeeds and ignores the store image"
+else
+    fail "older installer copy failed: $(cat "$TMP/old.err")"
+fi
+mkdir -p "$TMP/nested/microvm/sqs-store" "$TMP/nested-dest"
+printf 'x\n' > "$TMP/nested/microvm/vmlinuz"
+if cp "$TMP/nested/microvm/"* "$TMP/nested-dest/" 2> /dev/null; then
+    fail "control: a nested directory did not break the older copy, so the test proves nothing"
+else
+    pass "control: a nested directory would have broken the older copy"
+fi
+
+# --- install ---
+install_image "$TMP/rel" "$TMP/mv"
+if [ "$RC" = 0 ] && same "$TMP/built/vmlinuz" "$TMP/mv/sqs-store/vmlinuz" \
+    && same "$TMP/built/initramfs.cpio.gz" "$TMP/mv/sqs-store/initramfs.cpio.gz"; then
+    pass "installs the staged image where the daemon looks"
+else
+    fail "install rc=$RC: $OUT"
+fi
+mode=$(stat -c '%a' "$TMP/mv/sqs-store/initramfs.cpio.gz" 2> /dev/null || echo none)
+if [ "$mode" = 644 ]; then pass "installs the image 0644"; else fail "initramfs mode $mode"; fi
+
+image "$TMP/built2" v2
+release "$TMP/rel2"
+sh "$STAGE" "$TMP/built2" "$TMP/rel2" > /dev/null
+install_image "$TMP/rel2" "$TMP/mv"
+if [ "$RC" = 0 ] && same "$TMP/built2/vmlinuz" "$TMP/mv/sqs-store/vmlinuz" \
+    && same "$TMP/built2/initramfs.cpio.gz" "$TMP/mv/sqs-store/initramfs.cpio.gz"; then
+    pass "an upgrade replaces both files"
+else
+    fail "upgrade rc=$RC: $OUT"
+fi
+
+# A release without the store image: nothing installed, and a clear warning.
+install_image "$TMP/rel-none" "$TMP/mv-none"
+if [ "$RC" = 0 ] && [ ! -e "$TMP/mv-none/sqs-store" ] && echo "$OUT" | grep -q 'no SQS store microVM image; SQS store launches will be refused'; then
+    pass "a release without the image installs nothing and warns"
+else
+    fail "release without image: rc=$RC exists=$([ -e "$TMP/mv-none/sqs-store" ] && echo yes || echo no) $OUT"
+fi
+
+# installed <dir>: a store image from an earlier install, plus an unrelated file.
+installed() {
+    image "$1/sqs-store" old
+    printf 'keep\n' > "$1/sqs-store/operator-note"
+}
+gone() { [ ! -e "$1/sqs-store/vmlinuz" ] && [ ! -e "$1/sqs-store/initramfs.cpio.gz" ]; }
+kept() { [ "$(cat "$1/sqs-store/operator-note" 2> /dev/null)" = keep ]; }
+
+# Upgrading to a release without the image must not leave the old one usable.
+installed "$TMP/mv-up"
+install_image "$TMP/rel-none" "$TMP/mv-up"
+if [ "$RC" = 0 ] && gone "$TMP/mv-up" && kept "$TMP/mv-up" \
+    && echo "$OUT" | grep -q 'Removed the SQS store microVM image left by an earlier install' \
+    && echo "$OUT" | grep -q 'SQS store launches will be refused'; then
+    pass "a release without the image removes the earlier one, keeps other files, and warns"
+else
+    fail "release without image over an installed one: rc=$RC $OUT"
+fi
+
+# An incomplete store image refuses and removes the earlier pair, because the
+# binaries this release installed before it would otherwise boot the old one.
+installed "$TMP/mv-bad"
+cp -r "$TMP/rel2" "$TMP/rel-bad"
+rm "$TMP/rel-bad/microvm-sqs-store/initramfs.cpio.gz"
+install_image "$TMP/rel-bad" "$TMP/mv-bad"
+if [ "$RC" != 0 ] && echo "$OUT" | grep -q 'incomplete: initramfs.cpio.gz is missing' \
+    && gone "$TMP/mv-bad" && kept "$TMP/mv-bad"; then
+    pass "an incomplete image is refused and the earlier pair is removed"
+else
+    fail "incomplete image: rc=$RC $OUT"
+fi
+
+# Restaging from a missing or incomplete build drops an image staged earlier.
+for case in missing incomplete; do
+    image "$TMP/built-$case" new
+    release "$TMP/rel-$case"
+    sh "$STAGE" "$TMP/built" "$TMP/rel-$case" > /dev/null
+    if [ "$case" = missing ]; then rm -r "$TMP/built-$case"; else rm "$TMP/built-$case/initramfs.cpio.gz"; fi
+    OUT=$(sh "$STAGE" "$TMP/built-$case" "$TMP/rel-$case")
+    if [ ! -e "$TMP/rel-$case/microvm-sqs-store" ] && echo "$OUT" | grep -q 'tarball will not include it'; then
+        pass "restaging from a $case build ships no earlier image"
+    else
+        fail "restaging from a $case build left: $(ls -R "$TMP/rel-$case/microvm-sqs-store" 2> /dev/null)"
+    fi
+done
+mkdir -p "$TMP/rel-other/microvm-sqs-store"
+image "$TMP/rel-other/microvm-sqs-store" old
+printf 'keep\n' > "$TMP/rel-other/microvm-sqs-store/note"
+sh "$STAGE" "$TMP/no-such-build" "$TMP/rel-other" > /dev/null
+if [ ! -e "$TMP/rel-other/microvm-sqs-store/vmlinuz" ] && [ "$(cat "$TMP/rel-other/microvm-sqs-store/note")" = keep ]; then
+    pass "restaging removes only the image files"
+else
+    fail "restaging touched more than the image files"
+fi
+
+# An install that stops at the initramfs leaves no pair the daemon would boot.
+mkdir -p "$TMP/bin"
+cat > "$TMP/bin/install" << 'EOF'
+#!/bin/sh
+for a in "$@"; do last=$a; done
+case "$last" in */initramfs.cpio.gz) echo "install: stopped" >&2; exit 1 ;; esac
+exec /usr/bin/install "$@"
+EOF
+chmod +x "$TMP/bin/install"
+install_image "$TMP/rel" "$TMP/mv" "$TMP/bin:"
+if [ "$RC" != 0 ] && same "$TMP/built/vmlinuz" "$TMP/mv/sqs-store/vmlinuz" && [ ! -e "$TMP/mv/sqs-store/initramfs.cpio.gz" ]; then
+    pass "an interrupted install leaves the initramfs missing, not mismatched"
+else
+    fail "interrupted install: rc=$RC initramfs=$([ -e "$TMP/mv/sqs-store/initramfs.cpio.gz" ] && echo present || echo missing)"
+fi
+
+[ "$fails" = 0 ] || { echo "FAIL: $fails"; exit 1; }
+echo "  setup_sqs_store_image_test ok"
@@ -237,7 +237,7 @@ fi
 log "stopping services"
 run sudo systemctl stop spinifex.target 2>/dev/null || true
 if ! $DRY_RUN; then
-    for unit in spinifex-predastore spinifex-viperblock spinifex-sqs-store spinifex-nats; do
+    for unit in spinifex-predastore spinifex-viperblock spinifex-nats; do
        if systemctl is-active --quiet "$unit"; then
            sudo systemctl stop "$unit" || true
        fi
@@ -362,9 +362,10 @@ rm_path /etc/systemd/system/openvswitch-switch.service.d/spinifex-perms.conf \
        /etc/systemd/system/ovn-controller.service.d/log-level.conf \
        /etc/systemd/system/openvswitch-ipsec.service.d/spinifex-perms.conf \
        /etc/systemd/system/ovn-northd.service.d/spinifex-perms.conf \
-        /etc/systemd/system/ovn-northd.service.d/no-manage-ovsdb.conf
+        /etc/systemd/system/ovn-northd.service.d/no-manage-ovsdb.conf \
+        /etc/systemd/system/ipsec.service.d/spinifex-guard.conf
 # Only if we emptied them — another package may own the same override dir.
-for d in openvswitch-switch ovs-vswitchd ovn-controller openvswitch-ipsec ovn-northd; do
+for d in openvswitch-switch ovs-vswitchd ovn-controller openvswitch-ipsec ovn-northd ipsec; do
    dir="/etc/systemd/system/${d}.service.d"
    if sudo test -d "$dir" && [ -z "$(sudo ls -A "$dir" 2>/dev/null)" ]; then
        rm_path "$dir"
@@ -431,6 +432,11 @@ if [ -z "$PLUGINDIR" ]; then
 fi
 [ -n "$PLUGINDIR" ] && rm_path "$PLUGINDIR/nbdkit-viperblock-plugin.so"

+# A leftover guard table drops all Geneve from this host's encap address until
+# reboot, and the helper is the one place that names it.
+if [ -x /usr/local/lib/spinifex/spinifex-set-ipsec-state ]; then
+    run sudo /usr/local/lib/spinifex/spinifex-set-ipsec-state unguard || true
+fi
 rm_path /usr/local/bin/spx \
        /usr/local/share/spinifex \
        /usr/local/lib/spinifex \
@@ -493,7 +499,7 @@ fi
 # userdel warns about a missing home. Order here is cosmetic, not correctness.
 # ---------------------------------------------------------------------------
 log "removing service users and groups"
-for svc in nats sqsstore gw daemon storage northstar viperblock vpcd ui; do
+for svc in nats gw daemon storage northstar viperblock vpcd ui; do
    user="spinifex-${svc}"
    if id -u "$user" >/dev/null 2>&1; then
        run sudo userdel "$user" 2>/dev/null || log "  NOTE: could not remove user $user"
@@ -619,6 +625,7 @@ if ! $DRY_RUN; then
             /etc/NetworkManager/conf.d/*spinifex* \
             /etc/systemd/system/*.service.d/spinifex-perms.conf \
             /etc/systemd/system/ovn-northd.service.d/no-manage-ovsdb.conf \
+             /etc/systemd/system/ipsec.service.d/spinifex-guard.conf \
             /etc/systemd/system/ovn-controller.service.d/log-level.conf; do
        exists "$p" && LEFT+=("$p")
    done
@@ -628,7 +635,7 @@ if ! $DRY_RUN; then
    for u in "${RESTART_UNITS[@]}"; do
        systemctl is-failed --quiet "$u" 2>/dev/null && LEFT+=("${u%.service}.service is failed")
    done
-    for svc in nats sqsstore gw daemon storage northstar viperblock vpcd ui; do
+    for svc in nats gw daemon storage northstar viperblock vpcd ui; do
        id -u "spinifex-${svc}" >/dev/null 2>&1 && LEFT+=("user spinifex-${svc}")
    done
    if command -v nft >/dev/null 2>&1 && sudo nft list table inet spinifex_filter >/dev/null 2>&1; then
@@ -3,6 +3,7 @@ package quota
 import (
 	"context"
 	"errors"
+	"log/slog"

 	"github.com/aws/aws-sdk-go/service/ec2"
 	"github.com/aws/aws-sdk-go/service/elbv2"
@@ -113,3 +114,23 @@ func (s *Service) EnforceRDSInstances(ctx context.Context, natsConn *nats.Conn,
 	}
 	return exceeds(len(out.DBInstances), want, limits.RDSInstances)
 }
+
+// LoadBalancerLimit is the load balancer cap DescribeAccountLimits reports.
+func (s *Service) LoadBalancerLimit(ctx context.Context, accountID string) (int, error) {
+	// An exempt account has no cap enforced against it, so the AWS default is a
+	// truer answer than a configured limit nothing consults.
+	if s.Exempt(accountID) {
+		return elbv2awsapi.DefaultLoadBalancerLimit, nil
+	}
+	_, limits, err := s.GetAccountQuota(ctx, accountID)
+	if err != nil {
+		if awserrors.IsErrorCode(err, awserrors.ErrorBlocked) {
+			return 0, err
+		}
+		slog.ErrorContext(ctx, "DescribeAccountLimits: failed to read account quota", "accountID", accountID, "err", err)
+		return 0, errors.New(awserrors.ErrorInternalError)
+	}
+	return limits.LoadBalancers, nil
+}
+
+var _ elbv2awsapi.Quota = (*Service)(nil)
@@ -6,6 +6,7 @@ import (
 	"math"
 	"testing"

+	elbv2awsapi "github.com/mulgadc/spinifex/spinifex/domains/elbv2/awsapi"
 	"github.com/mulgadc/spinifex/spinifex/foundation/aws/errors"
 )

@@ -82,3 +83,25 @@ func TestEnforceLiveExemptShortCircuits(t *testing.T) {
 		})
 	}
 }
+
+func TestLoadBalancerLimit_ExemptAccountReportsAWSDefault(t *testing.T) {
+	var disabled *Service
+	got, err := disabled.LoadBalancerLimit(t.Context(), "123456789012")
+	if err != nil {
+		t.Fatalf("LoadBalancerLimit: %v", err)
+	}
+	if got != elbv2awsapi.DefaultLoadBalancerLimit {
+		t.Errorf("limit = %d, want the AWS default %d", got, elbv2awsapi.DefaultLoadBalancerLimit)
+	}
+}
+
+func TestLoadBalancerLimit_EnabledAccountReportsConfiguredLimit(t *testing.T) {
+	svc := New(Limits{Enabled: true, LoadBalancers: 7}, nil)
+	got, err := svc.LoadBalancerLimit(t.Context(), "123456789012")
+	if err != nil {
+		t.Fatalf("LoadBalancerLimit: %v", err)
+	}
+	if got != 7 {
+		t.Errorf("limit = %d, want 7", got)
+	}
+}
@@ -5,6 +5,7 @@ import (
 	"testing"

 	"github.com/mulgadc/spinifex/internal/testkit"
+	"github.com/mulgadc/spinifex/spinifex/foundation/aws/errors"
 	"github.com/nats-io/nats.go/jetstream"
 	"github.com/stretchr/testify/require"
 )
@@ -134,3 +135,36 @@ func TestPutAccountQuotaWithoutBucket(t *testing.T) {
 	svc := New(baseLimits(), nil)
 	require.Error(t, svc.PutAccountQuota(t.Context(), "000000000002", Overrides{VCPUs: ptr(32)}, "operator"))
 }
+
+// The cap DescribeAccountLimits reports follows the account's stored override
+// and falls back to the configured baseline for an account without one.
+func TestLoadBalancerLimit_FollowsPerAccountOverride(t *testing.T) {
+	svc, _ := newOverrideService(t)
+	const upgraded, plain = "000000000002", "000000000003"
+	require.NoError(t, svc.PutAccountQuota(t.Context(), upgraded, Overrides{LoadBalancers: ptr(31)}, "operator"))
+
+	got, err := svc.LoadBalancerLimit(t.Context(), upgraded)
+	require.NoError(t, err)
+	require.Equal(t, 31, got)
+
+	got, err = svc.LoadBalancerLimit(t.Context(), plain)
+	require.NoError(t, err)
+	require.Equal(t, baseLimits().LoadBalancers, got)
+}
+
+// A sealed account is refused with Blocked, and an unreadable record is an
+// InternalError, never a made-up number.
+func TestLoadBalancerLimit_RefusesSealedAndUnreadableRecords(t *testing.T) {
+	svc, kv := newOverrideService(t)
+	const sealed, corrupt = "000000000004", "000000000005"
+	_, err := kv.Put(t.Context(), sealed, []byte(`{"state":"sealed"}`))
+	require.NoError(t, err)
+	_, err = kv.Put(t.Context(), corrupt, []byte(`not json`))
+	require.NoError(t, err)
+
+	_, err = svc.LoadBalancerLimit(t.Context(), sealed)
+	require.EqualError(t, err, awserrors.ErrorBlocked)
+
+	_, err = svc.LoadBalancerLimit(t.Context(), corrupt)
+	require.EqualError(t, err, awserrors.ErrorInternalError)
+}
@@ -2,9 +2,9 @@ package image

 import (
 	"context"
-	"errors"
 	"strings"

+	"github.com/aws/aws-sdk-go/aws"
 	"github.com/aws/aws-sdk-go/service/ec2"
 	ec2image "github.com/mulgadc/spinifex/spinifex/domains/ec2/image"
 	"github.com/mulgadc/spinifex/spinifex/foundation/aws/errors"
@@ -27,74 +27,86 @@ var validRegisterImageBootModes = map[string]bool{
 }

 // ValidateRegisterImageInput rejects inputs that ask for behaviour spinifex
-// doesn't have (PV, S3 bundles, kernel/ramdisk, IMDS/TPM/ENA hints) rather
-// than silently discarding them. Registration is pointer-only: caller supplies
-// a snapshot and we write a config.json that references it.
+// doesn't have (PV, S3 bundles, kernel/ramdisk, IMDS/TPM hints) rather than
+// silently discarding them. Registration is pointer-only: caller supplies a
+// snapshot and we write a config.json that references it.
 func ValidateRegisterImageInput(input *ec2.RegisterImageInput) error {
 	if input == nil {
-		return errors.New(awserrors.ErrorMissingParameter)
+		return awserrors.Errorf(awserrors.ErrorMissingParameter, "The request must contain the parameter Name")
 	}

 	if input.Name == nil || *input.Name == "" {
-		return errors.New(awserrors.ErrorMissingParameter)
+		return missingParam("Name")
 	}
 	if n := len(*input.Name); n < 3 || n > 128 {
-		return errors.New(awserrors.ErrorInvalidAMINameMalformed)
+		return invalidParam(awserrors.ErrorInvalidAMINameMalformed, "Name", *input.Name, "AMI names must be between 3 and 128 characters long.")
 	}

-	if input.ImageLocation != nil && *input.ImageLocation != "" {
-		return errors.New(awserrors.ErrorInvalidParameterValue)
+	if aws.StringValue(input.ImageLocation) != "" {
+		return unsupportedParam("ImageLocation", *input.ImageLocation)
 	}

-	if input.Architecture != nil && *input.Architecture != "" {
-		if !validRegisterImageArchitectures[*input.Architecture] {
-			return errors.New(awserrors.ErrorInvalidParameterValue)
-		}
+	if arch := aws.StringValue(input.Architecture); arch != "" && !validRegisterImageArchitectures[arch] {
+		return invalidParam(awserrors.ErrorInvalidParameterValue, "Architecture", arch, "Supported values are x86_64, arm64 and i386.")
 	}

-	if input.VirtualizationType != nil && *input.VirtualizationType != "" && *input.VirtualizationType != "hvm" {
-		return errors.New(awserrors.ErrorInvalidParameterValue)
+	if virt := aws.StringValue(input.VirtualizationType); virt != "" && virt != "hvm" {
+		return invalidParam(awserrors.ErrorInvalidParameterValue, "VirtualizationType", virt, "Only hvm is supported.")
 	}

-	if input.BootMode != nil && *input.BootMode != "" {
-		if !validRegisterImageBootModes[*input.BootMode] {
-			return errors.New(awserrors.ErrorInvalidParameterValue)
-		}
+	if mode := aws.StringValue(input.BootMode); mode != "" && !validRegisterImageBootModes[mode] {
+		return invalidParam(awserrors.ErrorInvalidParameterValue, "BootMode", mode, "Supported values are bios, uefi and uefi-preferred.")
 	}
-	if input.KernelId != nil && *input.KernelId != "" {
-		return errors.New(awserrors.ErrorInvalidParameterValue)
+	if aws.StringValue(input.KernelId) != "" {
+		return unsupportedParam("KernelId", *input.KernelId)
 	}
-	if input.RamdiskId != nil && *input.RamdiskId != "" {
-		return errors.New(awserrors.ErrorInvalidParameterValue)
+	if aws.StringValue(input.RamdiskId) != "" {
+		return unsupportedParam("RamdiskId", *input.RamdiskId)
 	}
-	if input.TpmSupport != nil && *input.TpmSupport != "" {
-		return errors.New(awserrors.ErrorInvalidParameterValue)
+	if aws.StringValue(input.TpmSupport) != "" {
+		return unsupportedParam("TpmSupport", *input.TpmSupport)
 	}
-	if input.ImdsSupport != nil && *input.ImdsSupport != "" {
-		return errors.New(awserrors.ErrorInvalidParameterValue)
+	if aws.StringValue(input.ImdsSupport) != "" {
+		return unsupportedParam("ImdsSupport", *input.ImdsSupport)
 	}
-	if input.EnaSupport != nil {
-		return errors.New(awserrors.ErrorInvalidParameterValue)
-	}
-	if input.SriovNetSupport != nil && *input.SriovNetSupport != "" {
-		return errors.New(awserrors.ErrorInvalidParameterValue)
+	if aws.StringValue(input.SriovNetSupport) != "" {
+		return unsupportedParam("SriovNetSupport", *input.SriovNetSupport)
 	}

 	if len(input.BlockDeviceMappings) == 0 {
-		return errors.New(awserrors.ErrorMissingParameter)
+		return missingParam("BlockDeviceMapping")
 	}

 	root := selectRootBlockDeviceMapping(input.BlockDeviceMappings, input.RootDeviceName)
 	if root == nil || root.Ebs == nil || root.Ebs.SnapshotId == nil || *root.Ebs.SnapshotId == "" {
-		return errors.New(awserrors.ErrorMissingParameter)
+		if name := aws.StringValue(input.RootDeviceName); name != "" && root == nil {
+			return awserrors.Errorf(awserrors.ErrorMissingParameter,
+				"Value (%s) for parameter RootDeviceName is invalid. No block device mapping has that DeviceName.", name)
+		}
+		return missingParam("SnapshotId in the root block device mapping")
 	}
 	if !strings.HasPrefix(*root.Ebs.SnapshotId, "snap-") {
-		return errors.New(awserrors.ErrorInvalidSnapshotIDMalformed)
+		return invalidParam(awserrors.ErrorInvalidSnapshotIDMalformed, "SnapshotId", *root.Ebs.SnapshotId, "Expected: 'snap-...'.")
 	}

 	return nil
 }

+// missingParam is EC2's wording for an absent required parameter.
+func missingParam(name string) error {
+	return awserrors.Errorf(awserrors.ErrorMissingParameter, "The request must contain the parameter %s", name)
+}
+
+// invalidParam is EC2's wording for a parameter value it refuses.
+func invalidParam(code, name, value, reason string) error {
+	return awserrors.Errorf(code, "Value (%s) for parameter %s is invalid. %s", value, name, reason)
+}
+
+// unsupportedParam refuses a parameter spinifex has no behaviour for.
+func unsupportedParam(name, value string) error {
+	return invalidParam(awserrors.ErrorInvalidParameterValue, name, value, name+" is not supported.")
+}
+
 // selectRootBlockDeviceMapping returns the BDM matching RootDeviceName if set,
 // else the first BDM carrying an EBS snapshot reference.
 func selectRootBlockDeviceMapping(mappings []*ec2.BlockDeviceMapping, rootDeviceName *string) *ec2.BlockDeviceMapping {
@@ -8,6 +8,7 @@ import (
 	"github.com/aws/aws-sdk-go/service/ec2"
 	"github.com/mulgadc/spinifex/spinifex/foundation/aws/errors"
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )

 func validRegisterImageInput() *ec2.RegisterImageInput {
@@ -33,30 +34,37 @@ func TestValidateRegisterImageInput(t *testing.T) {
 		input   *ec2.RegisterImageInput
 		wantErr bool
 		errMsg  string
+		// wantIn is the text the message must carry, so the caller can see which
+		// parameter was refused.
+		wantIn string
 	}{
 		{
 			name:    "NilInput",
 			input:   nil,
 			wantErr: true,
 			errMsg:  awserrors.ErrorMissingParameter,
+			wantIn:  "parameter Name",
 		},
 		{
 			name:    "MissingName",
 			mutate:  func(i *ec2.RegisterImageInput) { i.Name = nil },
 			wantErr: true,
 			errMsg:  awserrors.ErrorMissingParameter,
+			wantIn:  "parameter Name",
 		},
 		{
 			name:    "EmptyName",
 			mutate:  func(i *ec2.RegisterImageInput) { i.Name = aws.String("") },
 			wantErr: true,
 			errMsg:  awserrors.ErrorMissingParameter,
+			wantIn:  "parameter Name",
 		},
 		{
 			name:    "ShortName",
 			mutate:  func(i *ec2.RegisterImageInput) { i.Name = aws.String("ab") },
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidAMINameMalformed,
+			wantIn:  "for parameter Name is invalid",
 		},
 		{
 			name: "LongName",
@@ -69,12 +77,14 @@ func TestValidateRegisterImageInput(t *testing.T) {
 			},
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidAMINameMalformed,
+			wantIn:  "for parameter Name is invalid",
 		},
 		{
 			name:    "MissingBlockDeviceMappings",
 			mutate:  func(i *ec2.RegisterImageInput) { i.BlockDeviceMappings = nil },
 			wantErr: true,
 			errMsg:  awserrors.ErrorMissingParameter,
+			wantIn:  "parameter BlockDeviceMapping",
 		},
 		{
 			name: "MissingSnapshotIdInRootBDM",
@@ -83,6 +93,7 @@ func TestValidateRegisterImageInput(t *testing.T) {
 			},
 			wantErr: true,
 			errMsg:  awserrors.ErrorMissingParameter,
+			wantIn:  "parameter SnapshotId",
 		},
 		{
 			name: "MalformedSnapshotId",
@@ -91,6 +102,7 @@ func TestValidateRegisterImageInput(t *testing.T) {
 			},
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidSnapshotIDMalformed,
+			wantIn:  "Value (not-a-snapshot) for parameter SnapshotId is invalid",
 		},
 		{
 			name: "RootDeviceNameWithNoMatchingBDM",
@@ -99,6 +111,7 @@ func TestValidateRegisterImageInput(t *testing.T) {
 			},
 			wantErr: true,
 			errMsg:  awserrors.ErrorMissingParameter,
+			wantIn:  "Value (/dev/xvda) for parameter RootDeviceName is invalid",
 		},
 		{
 			name: "InvalidArchitecture",
@@ -107,6 +120,7 @@ func TestValidateRegisterImageInput(t *testing.T) {
 			},
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidParameterValue,
+			wantIn:  "Value (riscv) for parameter Architecture is invalid",
 		},
 		{
 			name: "ParavirtualVirtualizationRejected",
@@ -115,6 +129,7 @@ func TestValidateRegisterImageInput(t *testing.T) {
 			},
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidParameterValue,
+			wantIn:  "Value (paravirtual) for parameter VirtualizationType is invalid",
 		},
 		{
 			name: "ImageLocationRejected",
@@ -123,6 +138,7 @@ func TestValidateRegisterImageInput(t *testing.T) {
 			},
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidParameterValue,
+			wantIn:  "for parameter ImageLocation is invalid. ImageLocation is not supported",
 		},
 		{
 			name:    "BootModeAcceptedBios",
@@ -144,42 +160,52 @@ func TestValidateRegisterImageInput(t *testing.T) {
 			mutate:  func(i *ec2.RegisterImageInput) { i.BootMode = aws.String("legacy-bios") },
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidParameterValue,
+			wantIn:  "Value (legacy-bios) for parameter BootMode is invalid",
 		},
 		{
 			name:    "KernelIdRejected",
 			mutate:  func(i *ec2.RegisterImageInput) { i.KernelId = aws.String("aki-123") },
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidParameterValue,
+			wantIn:  "Value (aki-123) for parameter KernelId is invalid. KernelId is not supported",
 		},
 		{
 			name:    "RamdiskIdRejected",
 			mutate:  func(i *ec2.RegisterImageInput) { i.RamdiskId = aws.String("ari-123") },
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidParameterValue,
+			wantIn:  "Value (ari-123) for parameter RamdiskId is invalid. RamdiskId is not supported",
 		},
 		{
 			name:    "TpmSupportRejected",
 			mutate:  func(i *ec2.RegisterImageInput) { i.TpmSupport = aws.String("v2.0") },
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidParameterValue,
+			wantIn:  "Value (v2.0) for parameter TpmSupport is invalid. TpmSupport is not supported",
 		},
 		{
 			name:    "ImdsSupportRejected",
 			mutate:  func(i *ec2.RegisterImageInput) { i.ImdsSupport = aws.String("v2.0") },
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidParameterValue,
+			wantIn:  "Value (v2.0) for parameter ImdsSupport is invalid. ImdsSupport is not supported",
 		},
 		{
-			name:    "EnaSupportRejected",
+			name:    "EnaSupportAcceptedTrue",
 			mutate:  func(i *ec2.RegisterImageInput) { i.EnaSupport = aws.Bool(true) },
-			wantErr: true,
-			errMsg:  awserrors.ErrorInvalidParameterValue,
+			wantErr: false,
+		},
+		{
+			name:    "EnaSupportAcceptedFalse",
+			mutate:  func(i *ec2.RegisterImageInput) { i.EnaSupport = aws.Bool(false) },
+			wantErr: false,
 		},
 		{
 			name:    "SriovNetSupportRejected",
 			mutate:  func(i *ec2.RegisterImageInput) { i.SriovNetSupport = aws.String("simple") },
 			wantErr: true,
 			errMsg:  awserrors.ErrorInvalidParameterValue,
+			wantIn:  "Value (simple) for parameter SriovNetSupport is invalid. SriovNetSupport is not supported",
 		},
 		{
 			name: "ValidWithExplicitArch",
@@ -230,8 +256,12 @@ func TestValidateRegisterImageInput(t *testing.T) {

 			err := ValidateRegisterImageInput(input)
 			if tt.wantErr {
-				assert.Error(t, err)
-				assert.Equal(t, tt.errMsg, err.Error())
+				require.Error(t, err)
+				code, msg, ok := awserrors.ResolveErrorDetail(err)
+				require.True(t, ok, "the error must carry a registered code")
+				assert.Equal(t, tt.errMsg, code)
+				assert.NotEmpty(t, msg, "the response must say what was refused")
+				assert.Contains(t, msg, tt.wantIn)
 			} else {
 				assert.NoError(t, err)
 			}
@@ -243,8 +273,10 @@ func TestRegisterImage_GatewayValidationFailureReturnsEarly(t *testing.T) {
 	// nil natsConn is fine here: validation rejects the input before any
 	// NATS round-trip is attempted, so we exercise the early-exit path.
 	_, err := RegisterImage(context.Background(), &ec2.RegisterImageInput{}, nil, "000000000001")
-	assert.Error(t, err)
-	assert.Equal(t, awserrors.ErrorMissingParameter, err.Error())
+	require.Error(t, err)
+	code, ok := awserrors.ResolveErrorCode(err)
+	require.True(t, ok)
+	assert.Equal(t, awserrors.ErrorMissingParameter, code)
 }

 func TestSelectRootBlockDeviceMapping_NoRootDeviceName(t *testing.T) {
@@ -104,6 +104,7 @@ type AMI struct {
 	VolumeSizeGiB   uint64            `json:"volume_size_gib"`
 	SnapshotID      string            `json:"snapshot_id,omitempty"`
 	BootMode        string            `json:"boot_mode,omitempty"`
+	EnaSupport      *bool             `json:"ena_support,omitempty"`
 	Distro          string            `json:"distro,omitempty"`
 	DistroFamily    string            `json:"distro_family,omitempty"`
 	Tags            map[string]string `json:"tags,omitempty"`
@@ -193,3 +193,32 @@ func TestCreateImageFromInstance_DoesNotInheritSourceDigest(t *testing.T) {
 	assert.Equal(t, "uefi", ami.BootMode, "the source AMI must have been read")
 	assert.Nil(t, ami.SourceDigest)
 }
+
+// AWS takes an image's EnaSupport from the instance it was made from, and every
+// instance here reports it, so the source AMI's value is not inherited.
+func TestCreateImageFromInstance_SetsEnaSupportFromInstance(t *testing.T) {
+	svc, _, provider := setupProviderSnapshotImageService(t)
+	doc := seedProviderVolume(t, svc, provider, "vol-ena001", 8)
+	require.NoError(t, svc.MetadataStore().PutAMI(context.Background(), ebsmetadata.AMI{
+		ImageID:         "ami-src-ena",
+		Name:            "no-ena",
+		ImageOwnerAlias: testAccountID,
+		EnaSupport:      aws.Bool(false),
+	}))
+
+	out, err := svc.CreateImageFromInstance(CreateImageParams{
+		Input: &ec2.CreateImageInput{
+			InstanceId: aws.String("i-ena001"),
+			Name:       aws.String("from-no-ena"),
+		},
+		RootVolumeID:  doc.VolumeID,
+		SourceImageID: "ami-src-ena",
+		IsRunning:     false,
+	}, testAccountID)
+	require.NoError(t, err)
+
+	ami, err := svc.MetadataStore().GetAMI(context.Background(), aws.StringValue(out.ImageId))
+	require.NoError(t, err)
+	require.NotNil(t, ami.EnaSupport)
+	assert.True(t, *ami.EnaSupport)
+}
@@ -324,6 +324,11 @@ func renderVisibleImage(amiMeta ebsmetadata.AMI, input *ec2.DescribeImagesInput,
 		image.BootMode = aws.String(amiMeta.BootMode)
 	}

+	// EnaSupport is optional on RegisterImage; AWS omits it when never set.
+	if amiMeta.EnaSupport != nil {
+		image.EnaSupport = aws.Bool(*amiMeta.EnaSupport)
+	}
+
 	if bdms := synthesizeRootBlockDeviceMapping(amiMeta, encryptedAtRest); bdms != nil {
 		image.RootDeviceName = aws.String("/dev/sda1")
 		image.BlockDeviceMappings = bdms
@@ -496,8 +501,10 @@ func (s *ImageServiceImpl) CreateImageFromInstance(params CreateImageParams, acc
 		RootDeviceType:  ec2.DeviceTypeEbs,
 		ImageOwnerAlias: accountID,
 		BootMode:        sourceAMI.BootMode,
-		Distro:          sourceAMI.Distro,
-		DistroFamily:    sourceAMI.DistroFamily,
+		// Every instance reports ENA support, and AWS takes the image's from the instance.
+		EnaSupport:   aws.Bool(true),
+		Distro:       sourceAMI.Distro,
+		DistroFamily: sourceAMI.DistroFamily,
 		// Snapshot succeeded before this point, so the image is complete.
 		State: "available",
 	}
@@ -901,6 +908,7 @@ func (s *ImageServiceImpl) CopyImage(ctx context.Context, input *ec2.CopyImageIn
 		ImageOwnerAlias: accountID,
 		CreationDate:    time.Now().UTC(),
 		BootMode:        srcMeta.BootMode,
+		EnaSupport:      copyBoolPtr(srcMeta.EnaSupport),
 		Distro:          srcMeta.Distro,
 		DistroFamily:    srcMeta.DistroFamily,
 		Tags:            tags,
@@ -1110,6 +1118,7 @@ func (s *ImageServiceImpl) RegisterImage(ctx context.Context, input *ec2.Registe
 		CreationDate:    time.Now().UTC(),
 		Tags:            tags,
 		BootMode:        aws.StringValue(input.BootMode),
+		EnaSupport:      copyBoolPtr(input.EnaSupport),
 		// The referenced snapshot already exists (checked above), so the
 		// image is complete as soon as this config is written.
 		State: "available",
@@ -1124,6 +1133,15 @@ func (s *ImageServiceImpl) RegisterImage(ctx context.Context, input *ec2.Registe
 	return &ec2.RegisterImageOutput{ImageId: aws.String(amiID)}, nil
 }

+// copyBoolPtr returns a pointer to a copy of *b, or nil, so two AMI records
+// never share one flag.
+func copyBoolPtr(b *bool) *bool {
+	if b == nil {
+		return nil
+	}
+	return aws.Bool(*b)
+}
+
 // pickRootSnapshotBDM returns the BDM matching RootDeviceName (if set) that
 // carries a non-empty EBS snapshot reference, else the first such BDM.
 func pickRootSnapshotBDM(mappings []*ec2.BlockDeviceMapping, rootDeviceName *string) *ec2.BlockDeviceMapping {
@@ -2,11 +2,13 @@ package image

 import (
 	"context"
+	"fmt"
 	awsidentifiers "github.com/mulgadc/spinifex/spinifex/foundation/aws/identifiers"
 	"io"
 	"log/slog"
 	"net/http"
 	"net/http/httptest"
+	"strconv"
 	"strings"
 	"testing"
 	"time"
@@ -1329,6 +1331,42 @@ func TestRegisterImage_HappyPath(t *testing.T) {
 	assert.Equal(t, testAccountID, *img.OwnerId)
 }

+func TestRegisterImage_EnaSupport(t *testing.T) {
+	tests := []struct {
+		name string
+		in   *bool
+		want *bool
+	}{
+		{"true is stored and described", aws.Bool(true), aws.Bool(true)},
+		{"false is stored and described", aws.Bool(false), aws.Bool(false)},
+		{"unset is omitted", nil, nil},
+	}
+	for i, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			svc, store := setupTestImageService(t)
+			snapID := fmt.Sprintf("snap-ena%03d", i)
+			putTestSnapshotConfig(t, store, snapID, 8, testAccountID)
+
+			input := validRegisterImageServiceInput(snapID)
+			input.EnaSupport = tt.in
+			out, err := svc.RegisterImage(context.Background(), input, testAccountID)
+			require.NoError(t, err)
+
+			desc, err := svc.DescribeImages(context.Background(), &ec2.DescribeImagesInput{
+				ImageIds: []*string{out.ImageId},
+			}, testAccountID)
+			require.NoError(t, err)
+			require.Len(t, desc.Images, 1)
+			if tt.want == nil {
+				assert.Nil(t, desc.Images[0].EnaSupport, "an image registered without EnaSupport must omit the member")
+				return
+			}
+			require.NotNil(t, desc.Images[0].EnaSupport)
+			assert.Equal(t, *tt.want, *desc.Images[0].EnaSupport)
+		})
+	}
+}
+
 func TestRegisterImage_DuplicateName(t *testing.T) {
 	svc, store := setupTestImageService(t)

@@ -1798,12 +1836,41 @@ func TestCopyImage_InheritsSourceFields(t *testing.T) {
 	assert.Equal(t, uint64(32), newMeta.VolumeSizeGiB)
 	assert.Equal(t, "ebs", newMeta.RootDeviceType)
 	assert.Equal(t, "uefi", newMeta.BootMode, "CopyImage must propagate BootMode from source")
+	assert.Nil(t, newMeta.EnaSupport, "an unset EnaSupport stays unset on the copy")
 	assert.Equal(t, "debian-12", newMeta.Distro, "CopyImage must propagate Distro from source")
 	assert.Equal(t, "debian", newMeta.DistroFamily, "cloud-init family detection reads DistroFamily")
 	assert.Equal(t, digest, newMeta.SourceDigest, "a zero-copy share still holds the source artifact")
 	assert.False(t, newMeta.CreationDate.Before(before), "CreationDate must be refreshed on copy, not inherited")
 }

+func TestCopyImage_CarriesEnaSupport(t *testing.T) {
+	for _, want := range []bool{true, false} {
+		t.Run(strconv.FormatBool(want), func(t *testing.T) {
+			svc, store := setupTestImageService(t)
+			putTestSnapshotDoc(t, store, ebsmetadata.Snapshot{
+				SnapshotID: "snap-ena-src", VolumeID: "vol-ena", VolumeSize: 8,
+				State: "completed", Progress: "100%", OwnerID: testAccountID,
+			})
+			putTestAMIConfigWithSnapshot(t, store, "ami-ena-src", "ena-source", testAccountID, "snap-ena-src", ebsmetadata.AMI{
+				Architecture:    "x86_64",
+				PlatformDetails: "Linux/UNIX",
+				Virtualization:  "hvm",
+				VolumeSizeGiB:   8,
+				RootDeviceType:  "ebs",
+				EnaSupport:      aws.Bool(want),
+			})
+
+			out, err := svc.CopyImage(context.Background(), validCopyImageServiceInput("ami-ena-src", "ena-copy"), testAccountID)
+			require.NoError(t, err)
+
+			newMeta, err := svc.GetAMIConfig(context.Background(), *out.ImageId)
+			require.NoError(t, err)
+			require.NotNil(t, newMeta.EnaSupport)
+			assert.Equal(t, want, *newMeta.EnaSupport)
+		})
+	}
+}
+
 func TestCopyImage_NewSnapshotSharesSourceVolumeID(t *testing.T) {
 	svc, store := setupTestImageService(t)

@@ -11,6 +11,11 @@ import (
 	"github.com/mulgadc/spinifex/spinifex/runtime/compute/vm"
 )

+// instanceEnaSupport is what every instance reports for ENA. Guests use virtio
+// NICs, and DescribeInstances, RunInstances and DescribeInstanceAttribute all
+// answer from this one value.
+const instanceEnaSupport = true
+
 // InstanceProjection carries the caller-scoped values ProjectInstance needs.
 // All fields are plain data so the projection stays pure and table-testable.
 type InstanceProjection struct {
@@ -169,6 +174,7 @@ func ProjectInstance(v *vm.VM, cfg InstanceProjection) (inst *ec2.Instance, stat
 	// aws_instance ever settled. Interfaces are copied rather than mutated
 	// because instanceCopy shares their pointers with the stored instance.
 	instanceCopy.SourceDestCheck = aws.Bool(true)
+	instanceCopy.EnaSupport = aws.Bool(instanceEnaSupport)
 	if len(instanceCopy.NetworkInterfaces) > 0 {
 		nics := make([]*ec2.InstanceNetworkInterface, 0, len(instanceCopy.NetworkInterfaces))
 		for _, nic := range instanceCopy.NetworkInterfaces {
@@ -309,6 +309,29 @@ func TestProjectInstance_EnclaveOptionsAlwaysDisabled(t *testing.T) {
 	assert.False(t, aws.BoolValue(stopped.EnclaveOptions.Enabled))
 }

+// TestProjectInstance_EnaSupportAlwaysTrue covers DescribeInstances agreeing
+// with DescribeInstanceAttribute, including for a stored instance that
+// predates the field and one that carries a stale false.
+func TestProjectInstance_EnaSupportAlwaysTrue(t *testing.T) {
+	stale := fullVM(vm.StateRunning)
+	stale.Instance.EnaSupport = aws.Bool(false)
+
+	tests := []struct {
+		name string
+		vm   *vm.VM
+		cfg  InstanceProjection
+	}{
+		{"running", fullVM(vm.StateRunning), runningCfg()},
+		{"stopped", fullVM(vm.StateStopped), stoppedCfg(80, "stopped")},
+		{"stale", stale, runningCfg()},
+	}
+	for _, tt := range tests {
+		got, _ := ProjectInstance(tt.vm, tt.cfg)
+		require.NotNil(t, got.EnaSupport, tt.name)
+		assert.True(t, *got.EnaSupport, tt.name)
+	}
+}
+
 func TestParseInstanceIDFilter(t *testing.T) {
 	tests := []struct {
 		name    string
@@ -365,6 +365,7 @@ func (s *InstanceServiceImpl) RunInstance(input *ec2.RunInstancesInput) (*vm.VM,
 	// them null, so a caller comparing the two read a disagreement.
 	ec2Instance.SetRootDeviceName(ebsmetadata.RootDeviceName)
 	ec2Instance.SetRootDeviceType(ec2.DeviceTypeEbs)
+	ec2Instance.SetEnaSupport(instanceEnaSupport)

 	// Stamp the metadata-options block so DescribeInstances reports the posture;
 	// the hop limit and the IMDSv1 opt-in are request-driven.
@@ -3347,7 +3348,7 @@ func (s *InstanceServiceImpl) DescribeInstanceAttribute(ctx context.Context, inp
 		output.EbsOptimized = &ec2.AttributeBooleanValue{Value: &val}

 	case ec2.InstanceAttributeNameEnaSupport:
-		val := true
+		val := instanceEnaSupport
 		output.EnaSupport = &ec2.AttributeBooleanValue{Value: &val}

 	case ec2.InstanceAttributeNameSourceDestCheck:
@@ -122,6 +122,20 @@ func TestRunInstance_RootDeviceFieldsPopulated(t *testing.T) {
 	assert.Equal(t, ec2.DeviceTypeEbs, aws.StringValue(ec2Instance.RootDeviceType))
 }

+func TestRunInstance_ReportsEnaSupport(t *testing.T) {
+	svc := &InstanceServiceImpl{
+		instanceTypes: map[string]*ec2.InstanceTypeInfo{"t3.micro": {InstanceType: aws.String("t3.micro")}},
+	}
+
+	_, ec2Instance, err := svc.RunInstance(&ec2.RunInstancesInput{
+		ImageId:      aws.String("ami-0abcdef1234567890"),
+		InstanceType: aws.String("t3.micro"),
+	})
+	require.NoError(t, err)
+	require.NotNil(t, ec2Instance.EnaSupport)
+	assert.True(t, *ec2Instance.EnaSupport)
+}
+
 func TestVolumeTagsFromSpec(t *testing.T) {
 	specs := []*ec2.TagSpecification{
 		{ResourceType: aws.String("instance"), Tags: []*ec2.Tag{{Key: aws.String("Name"), Value: aws.String("node")}}},
@@ -7,10 +7,9 @@ import (
 	"github.com/stretchr/testify/require"
 )

-// The two ED25519 fingerprints AWS itself returned on 2026-07-22 (account
-// 744090693897, us-east-1), one per write path. They are recorded observations,
-// not derivations, and they pin the rendering without needing the key material
-// that produced them.
+// The two ED25519 fingerprints AWS itself returned in us-east-1 on 2026-07-22, one
+// per write path. They are recorded observations, not derivations, and pin the
+// rendering without the key material that produced them.
 const (
 	awsImportedED25519Fingerprint = "8ZVMcTKqdR5hxMjZI3t6b09EPIOE5QA6G9/Z63ejNBM="
 	awsCreatedED25519Fingerprint  = "odzmoB4gyxi1jn9vGD4c+bOpt1XUX1JUyaHaWHHCO2w="
@@ -0,0 +1,38 @@
+package snapshot
+
+import (
+	"context"
+	"testing"
+
+	"github.com/aws/aws-sdk-go/aws"
+	"github.com/aws/aws-sdk-go/service/ec2"
+	"github.com/mulgadc/spinifex/spinifex/domains/ec2/ebs/metadata"
+	"github.com/mulgadc/spinifex/spinifex/foundation/aws/errors"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// A volume still being prepared has no settled contents, so it cannot be
+// captured; once available the same volume snapshots normally.
+func TestCreateSnapshot_RefusesACreatingVolume(t *testing.T) {
+	svc, store := setupTestSnapshotService(t)
+	ctx := context.Background()
+	seedProviderVolume(t, svc, "vol-preparing", 4)
+	seedVolumeDocument(t, store, ebsmetadata.Volume{
+		VolumeID: "vol-preparing", TenantID: testAccountID, CapacityGiB: 4,
+		State: "creating", AvailabilityZone: "us-east-1a",
+	})
+
+	_, err := svc.CreateSnapshot(ctx, &ec2.CreateSnapshotInput{VolumeId: aws.String("vol-preparing")}, testAccountID)
+	assert.EqualError(t, err, awserrors.ErrorIncorrectState)
+	snaps, err := svc.metadata.ListSnapshotsStrict(ctx, testAccountID)
+	require.NoError(t, err)
+	assert.Empty(t, snaps)
+
+	meta, err := svc.metadata.GetVolume(ctx, testAccountID, "vol-preparing")
+	require.NoError(t, err)
+	meta.State = "available"
+	require.NoError(t, svc.metadata.PutVolume(ctx, meta))
+	_, err = svc.CreateSnapshot(ctx, &ec2.CreateSnapshotInput{VolumeId: aws.String("vol-preparing")}, testAccountID)
+	require.NoError(t, err)
+}
@@ -204,6 +204,10 @@ func (s *SnapshotServiceImpl) CreateSnapshot(ctx context.Context, input *ec2.Cre
 	if volume.CapacityGiB == 0 {
 		return nil, errors.New(awserrors.ErrorServerInternal)
 	}
+	// A volume still being prepared has no settled contents to capture.
+	if volume.State == string(ebsprovider.VolumeStateCreating) {
+		return nil, errors.New(awserrors.ErrorIncorrectState)
+	}

 	// Flush the writes buffered by whichever node serves the volume, so the
 	// provider reads a current checkpoint. An attached volume that cannot be
@@ -86,6 +86,11 @@ type SystemInstanceInput struct {
 	// endpoint serves the profile's role credentials. Empty leaves the VM
 	// without a profile (the agent then needs static keys).
 	IamInstanceProfileArn string `json:"iam_instance_profile_arn,omitempty"`
+
+	// Kind selects a non-load-balancer direct-boot guest; empty is the load
+	// balancer. SQSStore carries the KindSQSStore guest's volume and key.
+	Kind     Kind           `json:"kind,omitempty"`
+	SQSStore *SQSStoreInput `json:"sqs_store,omitempty"`
 }

 // ExtraENIInput describes an additional pre-created ENI to attach to a system
@@ -0,0 +1,76 @@
+package systeminstance
+
+import (
+	"errors"
+	"fmt"
+	"regexp"
+
+	"github.com/nats-io/nkeys"
+)
+
+// Kind selects the guest a direct-boot system instance runs. The empty kind
+// is the load balancer; every other kind is launched only on its own subject.
+type Kind string
+
+// KindSQSStore is the dedicated SQS store guest on a persistent data volume.
+const KindSQSStore Kind = "sqs-store"
+
+// SQSStoreLaunchSubjectRoot is the subject root for SQS store launches. It is
+// versioned and separate from system.LaunchInstance so a daemon that predates
+// it has no subscriber and the launch fails instead of booting another guest.
+const SQSStoreLaunchSubjectRoot = "system.LaunchSQSStore.v1"
+
+// SQSStoreLaunchSubject is the queue subject for an instance type; with a node
+// ID it is that node's targeted subject.
+func SQSStoreLaunchSubject(instanceType, nodeID string) string {
+	if nodeID == "" {
+		return fmt.Sprintf("%s.%s", SQSStoreLaunchSubjectRoot, instanceType)
+	}
+	return fmt.Sprintf("%s.%s.%s", SQSStoreLaunchSubjectRoot, instanceType, nodeID)
+}
+
+// SQSStoreInput is what an SQS store guest needs beyond its network: an
+// existing, already-formatted system-account volume and the one NATS user
+// public key the store accepts. It never carries a private key or seed.
+type SQSStoreInput struct {
+	DataVolumeID string `json:"data_volume_id"`
+	ClientNKey   string `json:"client_nkey"`
+}
+
+// ErrInvalidSQSStoreInput is returned for a malformed SQS store launch.
+var ErrInvalidSQSStoreInput = errors.New("sysinstance: invalid SQS store launch")
+
+// volumeIDPattern is the platform's volume ID shape. The guest finds its disk
+// by the ID without its dash, which must fit the 20-byte virtio-blk serial.
+var volumeIDPattern = regexp.MustCompile(`^vol-[0-9a-f]{17}$`)
+
+// ValidateSQSStore checks an SQS store launch's shape. It reads nothing, so a
+// valid result says only that the volume and key are well formed.
+func (in *SystemInstanceInput) ValidateSQSStore() error {
+	switch {
+	case in.Kind != KindSQSStore:
+		return fmt.Errorf("%w: kind %q is not %q", ErrInvalidSQSStoreInput, in.Kind, KindSQSStore)
+	case in.BootMode != "" && in.BootMode != BootDirect:
+		return fmt.Errorf("%w: boot mode %q, the store boots direct only", ErrInvalidSQSStoreInput, in.BootMode)
+	case in.SQSStore == nil:
+		return fmt.Errorf("%w: no data volume or client key", ErrInvalidSQSStoreInput)
+	case !volumeIDPattern.MatchString(in.SQSStore.DataVolumeID):
+		return fmt.Errorf("%w: data volume ID %q is malformed", ErrInvalidSQSStoreInput, in.SQSStore.DataVolumeID)
+	case !nkeys.IsValidPublicUserKey(in.SQSStore.ClientNKey):
+		// The value is not echoed: a caller that sent a seed here must not
+		// have it written to a log.
+		return fmt.Errorf("%w: client key is not a NATS user public key", ErrInvalidSQSStoreInput)
+	case in.LBAgentEnv != "" || in.CACert != "" || in.Scheme != "":
+		return fmt.Errorf("%w: load balancer settings on an SQS store launch", ErrInvalidSQSStoreInput)
+	}
+	return nil
+}
+
+// RefuseNonLoadBalancerKind keeps every non-load-balancer guest off the
+// legacy launch path, which would boot the load balancer image instead.
+func (in *SystemInstanceInput) RefuseNonLoadBalancerKind() error {
+	if in.Kind != "" || in.SQSStore != nil {
+		return fmt.Errorf("%w: kind %q must use its own launch subject", ErrInvalidSQSStoreInput, in.Kind)
+	}
+	return nil
+}
@@ -0,0 +1,98 @@
+package systeminstance
+
+import (
+	"strings"
+	"testing"
+
+	identifiers "github.com/mulgadc/spinifex/spinifex/foundation/aws/identifiers"
+	"github.com/nats-io/nkeys"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func validSQSStoreInput(t *testing.T) (*SystemInstanceInput, []byte) {
+	t.Helper()
+	kp, err := nkeys.CreateUser()
+	require.NoError(t, err)
+	pub, err := kp.PublicKey()
+	require.NoError(t, err)
+	seed, err := kp.Seed()
+	require.NoError(t, err)
+	return &SystemInstanceInput{
+		InstanceType: "sys.micro",
+		Kind:         KindSQSStore,
+		SQSStore:     &SQSStoreInput{DataVolumeID: "vol-0123456789abcdef0", ClientNKey: pub},
+	}, seed
+}
+
+func TestValidateSQSStore(t *testing.T) {
+	in, _ := validSQSStoreInput(t)
+	require.NoError(t, in.ValidateSQSStore())
+
+	_, seed := validSQSStoreInput(t)
+	account, err := nkeys.CreateAccount()
+	require.NoError(t, err)
+	accountPub, err := account.PublicKey()
+	require.NoError(t, err)
+
+	cases := map[string]func(*SystemInstanceInput){
+		"no kind":          func(in *SystemInstanceInput) { in.Kind = "" },
+		"other kind":       func(in *SystemInstanceInput) { in.Kind = "lambda" },
+		"ami boot":         func(in *SystemInstanceInput) { in.BootMode = BootAMI },
+		"no store input":   func(in *SystemInstanceInput) { in.SQSStore = nil },
+		"no volume":        func(in *SystemInstanceInput) { in.SQSStore.DataVolumeID = "" },
+		"malformed volume": func(in *SystemInstanceInput) { in.SQSStore.DataVolumeID = "vol-XYZ" },
+		"no key":           func(in *SystemInstanceInput) { in.SQSStore.ClientNKey = "" },
+		"seed as key":      func(in *SystemInstanceInput) { in.SQSStore.ClientNKey = string(seed) },
+		"account key":      func(in *SystemInstanceInput) { in.SQSStore.ClientNKey = accountPub },
+		"lb agent env":     func(in *SystemInstanceInput) { in.LBAgentEnv = "LB_ID=x" },
+		"lb ca cert":       func(in *SystemInstanceInput) { in.CACert = "cert" },
+		"lb scheme":        func(in *SystemInstanceInput) { in.Scheme = "internal" },
+	}
+	for name, mutate := range cases {
+		t.Run(name, func(t *testing.T) {
+			in, _ := validSQSStoreInput(t)
+			mutate(in)
+			err := in.ValidateSQSStore()
+			require.ErrorIs(t, err, ErrInvalidSQSStoreInput)
+			assert.NotContains(t, err.Error(), string(seed), "a seed must never be echoed")
+		})
+	}
+}
+
+func TestValidateSQSStore_VolumeIDMatchesTheGuestSerial(t *testing.T) {
+	accepted := identifiers.GenerateResourceID("vol")
+	in, _ := validSQSStoreInput(t)
+	in.SQSStore.DataVolumeID = accepted
+	require.NoError(t, in.ValidateSQSStore(), "a platform-generated volume ID is accepted")
+	// The guest accepts only vol + 17 lowercase hex, the full 20-byte serial.
+	assert.Len(t, strings.ReplaceAll(accepted, "-", ""), 20)
+
+	for _, id := range []string{
+		// 16 hex
+		"vol-0123456789abcdef",
+		// 18 hex: serial over 20 bytes
+		"vol-0123456789abcdef01",
+		// 8 hex
+		"vol-01234567",
+		// upper case
+		"vol-0123456789ABCDEF0",
+		// no dash
+		"vol0123456789abcdef0",
+		// not a volume
+		"snap-0123456789abcdef0",
+		// trailing newline
+		"vol-0123456789abcdef0\n",
+	} {
+		in, _ := validSQSStoreInput(t)
+		in.SQSStore.DataVolumeID = id
+		require.ErrorIs(t, in.ValidateSQSStore(), ErrInvalidSQSStoreInput, id)
+	}
+}
+
+func TestRefuseNonLoadBalancerKind(t *testing.T) {
+	require.NoError(t, (&SystemInstanceInput{InstanceType: "sys.micro"}).RefuseNonLoadBalancerKind())
+	in, _ := validSQSStoreInput(t)
+	require.ErrorIs(t, in.RefuseNonLoadBalancerKind(), ErrInvalidSQSStoreInput)
+	require.Error(t, (&SystemInstanceInput{SQSStore: &SQSStoreInput{}}).RefuseNonLoadBalancerKind())
+}
@@ -0,0 +1,158 @@
+package volume
+
+//test:in-package — the creating state is asserted on the metadata document,
+// and the metadata store is unexported.
+
+import (
+	"context"
+	"testing"
+
+	"github.com/aws/aws-sdk-go/aws"
+	"github.com/aws/aws-sdk-go/service/ec2"
+	"github.com/mulgadc/spinifex/spinifex/foundation/aws/errors"
+	"github.com/mulgadc/spinifex/spinifex/providers/ebs"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func createPreparingVolume(t *testing.T, svc *VolumeServiceImpl) string {
+	t.Helper()
+	vol, err := svc.CreateVolumeForPreparation(context.Background(), &ec2.CreateVolumeInput{
+		Size: aws.Int64(4), AvailabilityZone: aws.String("ap-southeast-2a"),
+	}, testVolAccountID)
+	require.NoError(t, err)
+	assert.Equal(t, "creating", aws.StringValue(vol.State))
+	return aws.StringValue(vol.VolumeId)
+}
+
+func requireVolumeState(t *testing.T, svc *VolumeServiceImpl, volumeID, state string) {
+	t.Helper()
+	meta, err := svc.metadata.GetVolume(context.Background(), testVolAccountID, volumeID)
+	require.NoError(t, err)
+	assert.Equal(t, state, meta.State)
+	assert.Empty(t, meta.AttachedInstance)
+	assert.Equal(t, uint64(4), meta.CapacityGiB)
+}
+
+func TestCreateVolumeForPreparation_IsBornCreatingWhileCreateVolumeStaysAvailable(t *testing.T) {
+	svc := newTestVolumeService("ap-southeast-2a")
+	id := createPreparingVolume(t, svc)
+	requireVolumeState(t, svc, id, "creating")
+
+	described, err := svc.DescribeVolumes(context.Background(), &ec2.DescribeVolumesInput{VolumeIds: []*string{aws.String(id)}}, testVolAccountID)
+	require.NoError(t, err)
+	require.Len(t, described.Volumes, 1)
+	assert.Equal(t, "creating", aws.StringValue(described.Volumes[0].State))
+
+	normal, err := svc.CreateVolume(context.Background(), &ec2.CreateVolumeInput{
+		Size: aws.Int64(4), AvailabilityZone: aws.String("ap-southeast-2a"),
+	}, testVolAccountID)
+	require.NoError(t, err)
+	assert.Equal(t, "available", aws.StringValue(normal.State))
+	requireVolumeState(t, svc, aws.StringValue(normal.VolumeId), "available")
+}
+
+// Every public path that could take, change or capture the volume refuses it
+// and leaves it exactly as it was until preparation completes.
+func TestCreatingVolume_PublicMutationsAreRefused(t *testing.T) {
+	svc := newTestVolumeService("ap-southeast-2a")
+	svc.snapshotKV = setupTestVolumeKV(t)
+	id := createPreparingVolume(t, svc)
+	ctx := context.Background()
+
+	_, err := svc.DeleteVolume(ctx, &ec2.DeleteVolumeInput{VolumeId: aws.String(id)}, testVolAccountID)
+	assert.EqualError(t, err, awserrors.ErrorVolumeInUse)
+	requireVolumeState(t, svc, id, "creating")
+
+	_, err = svc.ModifyVolume(ctx, &ec2.ModifyVolumeInput{VolumeId: aws.String(id), Size: aws.Int64(8)}, testVolAccountID)
+	assert.EqualError(t, err, awserrors.ErrorIncorrectState)
+	requireVolumeState(t, svc, id, "creating")
+
+	_, err = svc.ForceDetachVolume(ctx, &ec2.DetachVolumeInput{VolumeId: aws.String(id), Force: aws.Bool(true)}, testVolAccountID)
+	assert.EqualError(t, err, awserrors.ErrorIncorrectState)
+	requireVolumeState(t, svc, id, "creating")
+}
+
+func TestCompleteVolumePreparation_ReleasesTheVolume(t *testing.T) {
+	svc := newTestVolumeService("ap-southeast-2a")
+	svc.snapshotKV = setupTestVolumeKV(t)
+	id := createPreparingVolume(t, svc)
+	ctx := context.Background()
+
+	require.NoError(t, svc.CompleteVolumePreparation(ctx, testVolAccountID, id))
+	requireVolumeState(t, svc, id, "available")
+
+	// Released means ordinary behavior again: modify and delete now proceed.
+	_, err := svc.ModifyVolume(ctx, &ec2.ModifyVolumeInput{VolumeId: aws.String(id), Size: aws.Int64(8)}, testVolAccountID)
+	require.NoError(t, err)
+	_, err = svc.DeleteVolume(ctx, &ec2.DeleteVolumeInput{VolumeId: aws.String(id)}, testVolAccountID)
+	require.NoError(t, err)
+}
+
+// Completion only releases an unattached creating volume, so it can never
+// flip a volume some other path owns to available.
+func TestCompleteVolumePreparation_RefusesAnythingButAnUnattachedCreatingVolume(t *testing.T) {
+	svc := newTestVolumeService("ap-southeast-2a")
+	ctx := context.Background()
+
+	normal, err := svc.CreateVolume(ctx, &ec2.CreateVolumeInput{
+		Size: aws.Int64(4), AvailabilityZone: aws.String("ap-southeast-2a"),
+	}, testVolAccountID)
+	require.NoError(t, err)
+	err = svc.CompleteVolumePreparation(ctx, testVolAccountID, aws.StringValue(normal.VolumeId))
+	require.ErrorContains(t, err, awserrors.ErrorIncorrectState)
+
+	require.NoError(t, svc.UpdateVolumeState(testVolAccountID, aws.StringValue(normal.VolumeId), "in-use", "i-0123456789abcdef0", "/dev/sdf"))
+	err = svc.CompleteVolumePreparation(ctx, testVolAccountID, aws.StringValue(normal.VolumeId))
+	require.ErrorContains(t, err, awserrors.ErrorIncorrectState)
+	meta, err := svc.metadata.GetVolume(ctx, testVolAccountID, aws.StringValue(normal.VolumeId))
+	require.NoError(t, err)
+	assert.Equal(t, "in-use", meta.State)
+
+	id := createPreparingVolume(t, svc)
+	err = svc.CompleteVolumePreparation(ctx, "000000000099", id)
+	assert.EqualError(t, err, awserrors.ErrorInvalidVolumeNotFound)
+	requireVolumeState(t, svc, id, "creating")
+
+	err = svc.CompleteVolumePreparation(ctx, testVolAccountID, "vol-0000000000000dead")
+	assert.EqualError(t, err, awserrors.ErrorInvalidVolumeNotFound)
+}
+
+func TestDiscardVolumePreparation_DeletesTheCreatingVolumeAndItsStorage(t *testing.T) {
+	svc := newTestVolumeService("ap-southeast-2a")
+	id := createPreparingVolume(t, svc)
+	ctx := context.Background()
+
+	require.NoError(t, svc.DiscardVolumePreparation(ctx, testVolAccountID, id))
+	_, err := svc.metadata.GetVolume(ctx, testVolAccountID, id)
+	require.Error(t, err, "the record must be gone")
+	_, err = svc.provider.GetVolume(ctx, ebsprovider.GetVolumeRequest{Versioned: ebsprovider.NewVersioned(), VolumeID: id})
+	require.ErrorIs(t, err, ebsprovider.ErrNotFound, "the provider volume must be gone")
+}
+
+// Discard is only for a failed preparation, so an ordinary or attached volume,
+// or one in another account, is never deleted through it.
+func TestDiscardVolumePreparation_RefusesAnythingButAnUnattachedCreatingVolume(t *testing.T) {
+	svc := newTestVolumeService("ap-southeast-2a")
+	ctx := context.Background()
+
+	normal, err := svc.CreateVolume(ctx, &ec2.CreateVolumeInput{
+		Size: aws.Int64(4), AvailabilityZone: aws.String("ap-southeast-2a"),
+	}, testVolAccountID)
+	require.NoError(t, err)
+	normalID := aws.StringValue(normal.VolumeId)
+	err = svc.DiscardVolumePreparation(ctx, testVolAccountID, normalID)
+	require.ErrorContains(t, err, awserrors.ErrorIncorrectState)
+	requireVolumeState(t, svc, normalID, "available")
+
+	require.NoError(t, svc.UpdateVolumeState(testVolAccountID, normalID, "in-use", "i-0123456789abcdef0", "/dev/sdf"))
+	err = svc.DiscardVolumePreparation(ctx, testVolAccountID, normalID)
+	require.ErrorContains(t, err, awserrors.ErrorIncorrectState)
+
+	id := createPreparingVolume(t, svc)
+	err = svc.DiscardVolumePreparation(ctx, "000000000099", id)
+	assert.EqualError(t, err, awserrors.ErrorInvalidVolumeNotFound)
+	requireVolumeState(t, svc, id, "creating")
+	_, err = svc.provider.GetVolume(ctx, ebsprovider.GetVolumeRequest{Versioned: ebsprovider.NewVersioned(), VolumeID: id})
+	require.NoError(t, err, "a refused discard must leave the storage")
+}
@@ -118,6 +118,70 @@ func NewVolumeServiceImplWithStore(cfg *config.Config, store objectstore.ObjectS
 // CreateVolume allocates a new EBS volume through the EBS provider and records
 // its control-plane metadata document.
 func (s *VolumeServiceImpl) CreateVolume(ctx context.Context, input *ec2.CreateVolumeInput, accountID string) (*ec2.Volume, error) {
+	return s.createVolume(ctx, input, accountID, ebsprovider.VolumeStateAvailable)
+}
+
+// CreateVolumeForPreparation is CreateVolume for an internal caller that must
+// write the volume's first contents before anyone else can use it. The record
+// is born "creating", which attach, delete, modify, force-detach and snapshot
+// all refuse, until CompleteVolumePreparation makes it available.
+func (s *VolumeServiceImpl) CreateVolumeForPreparation(ctx context.Context, input *ec2.CreateVolumeInput, accountID string) (*ec2.Volume, error) {
+	return s.createVolume(ctx, input, accountID, ebsprovider.VolumeStateCreating)
+}
+
+// CompleteVolumePreparation moves a volume made by CreateVolumeForPreparation
+// to available. Anything other than an unattached "creating" volume is
+// refused, so it cannot release a volume some other path is using.
+func (s *VolumeServiceImpl) CompleteVolumePreparation(ctx context.Context, accountID, volumeID string) error {
+	meta, err := s.getVolumeMetadata(ctx, accountID, volumeID)
+	if err != nil {
+		return err
+	}
+	if meta.TenantID != accountID {
+		return errors.New(awserrors.ErrorInvalidVolumeNotFound)
+	}
+	if meta.State != string(ebsprovider.VolumeStateCreating) || meta.AttachedInstance != "" {
+		return awserrors.Errorf(awserrors.ErrorIncorrectState, "volume %s is %q, not an unattached creating volume", volumeID, meta.State)
+	}
+	meta.State = string(ebsprovider.VolumeStateAvailable)
+	if err := s.metadata.PutVolume(ctx, meta); err != nil {
+		return awserrors.Errorf(awserrors.ErrorServerInternal, "write volume metadata %s: %v", volumeID, err)
+	}
+	slog.InfoContext(ctx, "Volume preparation completed", "volumeId", volumeID)
+	return nil
+}
+
+// DiscardVolumePreparation deletes a volume whose preparation failed. Only an
+// unattached "creating" volume in the caller's account qualifies; the caller
+// must already know no export of it can still be live.
+func (s *VolumeServiceImpl) DiscardVolumePreparation(ctx context.Context, accountID, volumeID string) error {
+	if err := s.requireProvider(ctx, "DiscardVolumePreparation"); err != nil {
+		return err
+	}
+	meta, err := s.getVolumeMetadata(ctx, accountID, volumeID)
+	if err != nil {
+		return err
+	}
+	if meta.TenantID != accountID {
+		return errors.New(awserrors.ErrorInvalidVolumeNotFound)
+	}
+	if meta.State != string(ebsprovider.VolumeStateCreating) || meta.AttachedInstance != "" {
+		return awserrors.Errorf(awserrors.ErrorIncorrectState, "volume %s is %q, not an unattached creating volume", volumeID, meta.State)
+	}
+	// Snapshots refuse a creating volume, so none can depend on this one.
+	if err := s.provider.DeleteVolume(ctx, ebsprovider.DeleteVolumeRequest{
+		Versioned: ebsprovider.NewVersioned(), VolumeID: volumeID, Handle: meta.ProviderHandle,
+	}); err != nil {
+		return awserrors.Errorf(awserrors.ErrorServerInternal, "delete volume %s: %v", volumeID, err)
+	}
+	if err := s.metadata.DeleteVolume(ctx, accountID, volumeID); err != nil && !objectstore.IsNoSuchKeyError(err) {
+		return awserrors.Errorf(awserrors.ErrorServerInternal, "delete volume metadata %s: %v", volumeID, err)
+	}
+	slog.InfoContext(ctx, "Volume preparation discarded", "volumeId", volumeID)
+	return nil
+}
+
+func (s *VolumeServiceImpl) createVolume(ctx context.Context, input *ec2.CreateVolumeInput, accountID string, initialState ebsprovider.VolumeState) (*ec2.Volume, error) {
 	if input == nil {
 		return nil, errors.New(awserrors.ErrorInvalidParameterValue)
 	}
@@ -241,7 +305,7 @@ func (s *VolumeServiceImpl) CreateVolume(ctx context.Context, input *ec2.CreateV
 		return nil, errors.New(awserrors.ErrorServerInternal)
 	}
 	if err := s.metadata.PutVolume(ctx, ebsmetadata.Volume{
-		VolumeID: volumeID, TenantID: accountID, CapacityGiB: uint64(size), State: string(ebsprovider.VolumeStateAvailable),
+		VolumeID: volumeID, TenantID: accountID, CapacityGiB: uint64(size), State: string(initialState),
 		CreatedAt: now, AvailabilityZone: *input.AvailabilityZone, VolumeType: volumeType,
 		IOPS: iops, Throughput: throughput, SnapshotID: snapshotID, Tags: tags, ProviderHandle: created.Handle,
 		Encrypted: encrypted,
@@ -254,7 +318,7 @@ func (s *VolumeServiceImpl) CreateVolume(ctx context.Context, input *ec2.CreateV
 	slog.InfoContext(ctx, "CreateVolume completed", "volumeId", volumeID, "size", size, "type", volumeType)

 	return &ec2.Volume{VolumeId: aws.String(volumeID), Size: aws.Int64(size), VolumeType: aws.String(volumeType),
-		State: aws.String("available"), AvailabilityZone: input.AvailabilityZone, CreateTime: aws.Time(now),
+		State: aws.String(string(initialState)), AvailabilityZone: input.AvailabilityZone, CreateTime: aws.Time(now),
 		Iops: aws.Int64(int64(iops)), Throughput: aws.Int64(int64(throughput)), Encrypted: aws.Bool(encrypted),
 		SnapshotId: snapshotIDOrNil(snapshotID), Tags: awstags.MapToEC2(tags)}, nil
 }
@@ -947,7 +1011,7 @@ func (s *VolumeServiceImpl) ModifyVolume(ctx context.Context, input *ec2.ModifyV
 	if input.Size == nil || *input.Size <= originalSize {
 		return nil, errors.New(awserrors.ErrorInvalidParameterValue)
 	}
-	if meta.AttachedInstance != "" && meta.State == "in-use" {
+	if (meta.AttachedInstance != "" && meta.State == "in-use") || meta.State == string(ebsprovider.VolumeStateCreating) {
 		return nil, errors.New(awserrors.ErrorIncorrectState)
 	}
 	originalType := meta.VolumeType
@@ -1070,6 +1134,11 @@ func (s *VolumeServiceImpl) ForceDetachVolume(ctx context.Context, input *ec2.De
 		return nil, errors.New(awserrors.ErrorInvalidVolumeNotFound)
 	}

+	// Clearing would also make a volume still being prepared available.
+	if meta.State == string(ebsprovider.VolumeStateCreating) {
+		return nil, errors.New(awserrors.ErrorIncorrectState)
+	}
+
 	previous := meta.AttachedInstance
 	if err := s.UpdateVolumeState(accountID, volumeID, "available", "", ""); err != nil {
 		slog.ErrorContext(ctx, "ForceDetachVolume: failed to clear attachment", "volumeId", volumeID, "err", err)
@@ -0,0 +1,117 @@
+package cluster
+
+import (
+	"encoding/json"
+	"errors"
+	"log/slog"
+	"net/http"
+
+	"github.com/nats-io/nats.go/jetstream"
+)
+
+// OIDCDiscoveryDocument is the subset of the OpenID Connect Discovery metadata
+// (RFC 8414) that AWS STS and the AWS SDKs consume to validate IRSA
+// ServiceAccount tokens. Served unauthenticated at the per-cluster issuer URL.
+type OIDCDiscoveryDocument struct {
+	Issuer                           string   `json:"issuer"`
+	JWKSURI                          string   `json:"jwks_uri"`
+	ResponseTypesSupported           []string `json:"response_types_supported"`
+	SubjectTypesSupported            []string `json:"subject_types_supported"`
+	IDTokenSigningAlgValuesSupported []string `json:"id_token_signing_alg_values_supported"`
+	ClaimsSupported                  []string `json:"claims_supported"`
+}
+
+// OIDCPublisher serves each EKS cluster's OIDC discovery document and JWKS from
+// the per-account bucket this package owns. The caller resolves the account and
+// cluster from its route and mounts the handlers unauthenticated.
+type OIDCPublisher struct {
+	js jetstream.JetStream
+}
+
+// NewOIDCPublisher constructs an OIDCPublisher over js.
+func NewOIDCPublisher(js jetstream.JetStream) *OIDCPublisher {
+	return &OIDCPublisher{js: js}
+}
+
+// accountKV opens the read-only per-account EKS bucket. It returns
+// jetstream.ErrBucketNotFound (unwrapped) when the account has no clusters,
+// which the handlers translate to 404.
+func (p *OIDCPublisher) accountKV(r *http.Request, accountID string) (jetstream.KeyValue, error) {
+	if p == nil || p.js == nil {
+		return nil, errors.New("EKS OIDC publisher JetStream not initialised")
+	}
+	return p.js.KeyValue(r.Context(), AccountBucketName(accountID))
+}
+
+// ServeDiscoveryDocument serves the cluster's OpenID discovery document. The
+// `issuer` it returns is the cluster's stored IssuerURL — byte-identical to the
+// apiserver's `iss` claim — so an AWS SDK that fetches this document while
+// registering the provider sees a matching issuer.
+func (p *OIDCPublisher) ServeDiscoveryDocument(w http.ResponseWriter, r *http.Request, accountID, clusterName string) {
+	kv, err := p.accountKV(r, accountID)
+	if err != nil {
+		if errors.Is(err, jetstream.ErrBucketNotFound) {
+			http.NotFound(w, r)
+			return
+		}
+		slog.Error("OIDC discovery: open account bucket", "accountID", accountID, "err", err)
+		http.Error(w, "internal error", http.StatusInternalServerError)
+		return
+	}
+
+	meta, err := GetClusterMeta(r.Context(), kv, clusterName)
+	if err != nil || meta == nil || meta.OIDCIssuer == "" {
+		http.NotFound(w, r)
+		return
+	}
+
+	writeOIDCJSON(w, OIDCDiscoveryDocument{
+		Issuer:                           meta.OIDCIssuer,
+		JWKSURI:                          meta.OIDCIssuer + "/keys",
+		ResponseTypesSupported:           []string{"id_token"},
+		SubjectTypesSupported:            []string{"public"},
+		IDTokenSigningAlgValuesSupported: []string{"ES256"},
+		ClaimsSupported:                  []string{"sub", "iss", "aud", "exp", "iat"},
+	})
+}
+
+// ServeJWKS serves the cluster's JWKS document (the public verification key) as
+// referenced by the discovery document's jwks_uri. The stored bytes are written
+// as they are, without validating that they parse as a JWKS.
+func (p *OIDCPublisher) ServeJWKS(w http.ResponseWriter, r *http.Request, accountID, clusterName string) {
+	kv, err := p.accountKV(r, accountID)
+	if err != nil {
+		if errors.Is(err, jetstream.ErrBucketNotFound) {
+			http.NotFound(w, r)
+			return
+		}
+		slog.Error("OIDC JWKS: open account bucket", "accountID", accountID, "err", err)
+		http.Error(w, "internal error", http.StatusInternalServerError)
+		return
+	}
+
+	entry, err := kv.Get(r.Context(), OIDCJWKSKey(clusterName))
+	if err != nil {
+		http.NotFound(w, r)
+		return
+	}
+
+	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("Cache-Control", "public, max-age=300")
+	if _, err := w.Write(entry.Value()); err != nil {
+		slog.Debug("OIDC JWKS: write response", "err", err)
+	}
+}
+
+func writeOIDCJSON(w http.ResponseWriter, v any) {
+	body, err := json.Marshal(v)
+	if err != nil {
+		http.Error(w, "internal error", http.StatusInternalServerError)
+		return
+	}
+	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("Cache-Control", "public, max-age=300")
+	if _, err := w.Write(body); err != nil {
+		slog.Debug("OIDC discovery: write response", "err", err)
+	}
+}
@@ -0,0 +1,60 @@
+package cluster
+
+import (
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"github.com/mulgadc/spinifex/internal/testkit"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+const (
+	oidcPubAccount = "000000000001"
+	oidcPubCluster = "demo"
+)
+
+func serveOIDC(t *testing.T, p *OIDCPublisher, jwks bool) *httptest.ResponseRecorder {
+	t.Helper()
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequest(http.MethodGet, "/", nil)
+	if jwks {
+		p.ServeJWKS(rec, req, oidcPubAccount, oidcPubCluster)
+	} else {
+		p.ServeDiscoveryDocument(rec, req, oidcPubAccount, oidcPubCluster)
+	}
+	return rec
+}
+
+func TestOIDCPublisher_UnwiredAnswers500(t *testing.T) {
+	for _, p := range []*OIDCPublisher{nil, NewOIDCPublisher(nil)} {
+		assert.Equal(t, http.StatusInternalServerError, serveOIDC(t, p, false).Code)
+		assert.Equal(t, http.StatusInternalServerError, serveOIDC(t, p, true).Code)
+	}
+}
+
+func TestOIDCPublisher_NoAccountBucketAnswers404(t *testing.T) {
+	_, nc, _ := testutil.StartTestJetStream(t)
+	p := NewOIDCPublisher(testutil.NewJetStream(t, nc))
+
+	assert.Equal(t, http.StatusNotFound, serveOIDC(t, p, false).Code)
+	assert.Equal(t, http.StatusNotFound, serveOIDC(t, p, true).Code)
+}
+
+// A stored JWKS is written back as it is, even when it is not valid JSON.
+func TestOIDCPublisher_ServesStoredJWKSVerbatim(t *testing.T) {
+	_, nc, _ := testutil.StartTestJetStream(t)
+	js := testutil.NewJetStream(t, nc)
+	kv, err := GetOrCreateAccountBucket(t.Context(), js, oidcPubAccount)
+	require.NoError(t, err)
+	_, err = kv.Put(t.Context(), OIDCJWKSKey(oidcPubCluster), []byte("not json"))
+	require.NoError(t, err)
+
+	rec := serveOIDC(t, NewOIDCPublisher(js), true)
+
+	assert.Equal(t, http.StatusOK, rec.Code)
+	assert.Equal(t, "not json", rec.Body.String())
+	assert.Equal(t, "application/json", rec.Header().Get("Content-Type"))
+	assert.Equal(t, "public, max-age=300", rec.Header().Get("Cache-Control"))
+}
@@ -42,7 +42,7 @@ const (
 	sourceActionTargetGroups
 )

-// Every action ELBv2_Request serves, with the resources AWS evaluates it
+// Every action the ELBv2 registration serves, with the resources AWS evaluates it
 // against. Exhaustive by contract: a completeness test compares this table with
 // the dispatch table in both directions, so an action cannot be added with a
 // silent account-wide grant.
@@ -0,0 +1,323 @@
+package awsapi
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"io"
+	"log/slog"
+	"maps"
+	"net/http"
+	"slices"
+
+	"github.com/aws/aws-sdk-go/service/elbv2"
+	"github.com/mulgadc/bluebottle/pkg/sigv4"
+	elbv2domain "github.com/mulgadc/spinifex/spinifex/domains/elbv2"
+	"github.com/mulgadc/spinifex/spinifex/foundation/aws/errors"
+	awsxml "github.com/mulgadc/spinifex/spinifex/foundation/aws/xml"
+	"github.com/mulgadc/spinifex/spinifex/ingress/aws/dispatch"
+	"github.com/mulgadc/spinifex/spinifex/ingress/aws/query"
+	"github.com/nats-io/nats.go"
+)
+
+// ServiceName is the SigV4 credential-scope service ELBv2 clients sign for.
+const ServiceName = "elasticloadbalancing"
+
+// InventoryName keys the operation inventory; it differs from the credential
+// scope because the inventory is named for the API generation.
+const InventoryName = "elasticloadbalancingv2"
+
+// Quota is the account load balancer cap ELBv2 consults. It is declared here
+// because domains/admission/quota imports this package to count live load
+// balancers, so this package cannot import it back.
+type Quota interface {
+	EnforceLoadBalancers(ctx context.Context, nc *nats.Conn, accountID string, want int) error
+	Exempt(accountID string) bool
+	// LoadBalancerLimit is the cap the account's DescribeAccountLimits reports.
+	LoadBalancerLimit(ctx context.Context, accountID string) (int, error)
+}
+
+// Deps are the composed capabilities the ELBv2 control plane dispatches to. A
+// nil NATS connection answers every action with ServerInternal and a nil Quota
+// enforces and reports the AWS default.
+type Deps struct {
+	NATS  *nats.Conn
+	Quota Quota
+}
+
+// action parses once so authorization and dispatch share one typed input.
+type action struct {
+	parse    func(map[string]string) (any, error)
+	dispatch func(ctx context.Context, name string, parsed any, d Deps, accountID string) ([]byte, error)
+}
+
+// handler creates a type-safe action. Parsing is separate from dispatch so the
+// resolver authorizes the exact input the handler executes. ELBv2 uses the
+// IAM-style XML envelope: <ActionResponse><ActionResult>...</ActionResult></ActionResponse>.
+func handler[In any](h func(context.Context, *In, Deps, string) (any, error)) action {
+	return action{
+		parse: func(q map[string]string) (any, error) {
+			input := new(In)
+			if err := query.QueryParamsToStruct(q, input); err != nil {
+				return nil, err
+			}
+			return input, nil
+		},
+		dispatch: func(ctx context.Context, name string, parsed any, d Deps, accountID string) ([]byte, error) {
+			input, ok := parsed.(*In)
+			if !ok {
+				return nil, errors.New(awserrors.ErrorInternalError)
+			}
+			output, err := h(ctx, input, d, accountID)
+			if err != nil {
+				return nil, err
+			}
+			payload := awsxml.QueryResponsePayload(name, output)
+			xmlOutput, err := awsxml.Marshal(payload)
+			if err != nil {
+				return nil, errors.New("failed to marshal response to XML")
+			}
+			return xmlOutput, nil
+		},
+	}
+}
+
+var actions = map[string]action{
+	"CreateLoadBalancer": handler(func(ctx context.Context, input *elbv2.CreateLoadBalancerInput, d Deps, accountID string) (any, error) {
+		if d.Quota != nil {
+			if err := d.Quota.EnforceLoadBalancers(ctx, d.NATS, accountID, 1); err != nil {
+				return nil, err
+			}
+		}
+		return CreateLoadBalancer(ctx, input, d.NATS, accountID)
+	}),
+	"DeleteLoadBalancer": handler(func(ctx context.Context, input *elbv2.DeleteLoadBalancerInput, d Deps, accountID string) (any, error) {
+		return DeleteLoadBalancer(ctx, input, d.NATS, accountID)
+	}),
+	"DescribeLoadBalancers": handler(func(ctx context.Context, input *elbv2.DescribeLoadBalancersInput, d Deps, accountID string) (any, error) {
+		return DescribeLoadBalancers(ctx, input, d.NATS, accountID)
+	}),
+	"CreateTargetGroup": handler(func(ctx context.Context, input *elbv2.CreateTargetGroupInput, d Deps, accountID string) (any, error) {
+		return CreateTargetGroup(ctx, input, d.NATS, accountID)
+	}),
+	"DeleteTargetGroup": handler(func(ctx context.Context, input *elbv2.DeleteTargetGroupInput, d Deps, accountID string) (any, error) {
+		return DeleteTargetGroup(ctx, input, d.NATS, accountID)
+	}),
+	"DescribeTargetGroups": handler(func(ctx context.Context, input *elbv2.DescribeTargetGroupsInput, d Deps, accountID string) (any, error) {
+		return DescribeTargetGroups(ctx, input, d.NATS, accountID)
+	}),
+	"RegisterTargets": handler(func(ctx context.Context, input *elbv2.RegisterTargetsInput, d Deps, accountID string) (any, error) {
+		return RegisterTargets(ctx, input, d.NATS, accountID)
+	}),
+	"DeregisterTargets": handler(func(ctx context.Context, input *elbv2.DeregisterTargetsInput, d Deps, accountID string) (any, error) {
+		return DeregisterTargets(ctx, input, d.NATS, accountID)
+	}),
+	"DescribeTargetHealth": handler(func(ctx context.Context, input *elbv2.DescribeTargetHealthInput, d Deps, accountID string) (any, error) {
+		return DescribeTargetHealth(ctx, input, d.NATS, accountID)
+	}),
+	"CreateListener": handler(func(ctx context.Context, input *elbv2.CreateListenerInput, d Deps, accountID string) (any, error) {
+		return CreateListener(ctx, input, d.NATS, accountID)
+	}),
+	"DeleteListener": handler(func(ctx context.Context, input *elbv2.DeleteListenerInput, d Deps, accountID string) (any, error) {
+		return DeleteListener(ctx, input, d.NATS, accountID)
+	}),
+	"ModifyListener": handler(func(ctx context.Context, input *elbv2.ModifyListenerInput, d Deps, accountID string) (any, error) {
+		return ModifyListener(ctx, input, d.NATS, accountID)
+	}),
+	"DescribeListeners": handler(func(ctx context.Context, input *elbv2.DescribeListenersInput, d Deps, accountID string) (any, error) {
+		return DescribeListeners(ctx, input, d.NATS, accountID)
+	}),
+	"CreateRule": handler(func(ctx context.Context, input *elbv2.CreateRuleInput, d Deps, accountID string) (any, error) {
+		return CreateRule(ctx, input, d.NATS, accountID)
+	}),
+	"ModifyRule": handler(func(ctx context.Context, input *elbv2.ModifyRuleInput, d Deps, accountID string) (any, error) {
+		return ModifyRule(ctx, input, d.NATS, accountID)
+	}),
+	"DeleteRule": handler(func(ctx context.Context, input *elbv2.DeleteRuleInput, d Deps, accountID string) (any, error) {
+		return DeleteRule(ctx, input, d.NATS, accountID)
+	}),
+	"DescribeRules": handler(func(ctx context.Context, input *elbv2.DescribeRulesInput, d Deps, accountID string) (any, error) {
+		return DescribeRules(ctx, input, d.NATS, accountID)
+	}),
+	"SetRulePriorities": handler(func(ctx context.Context, input *elbv2.SetRulePrioritiesInput, d Deps, accountID string) (any, error) {
+		return SetRulePriorities(ctx, input, d.NATS, accountID)
+	}),
+	"DescribeTags": handler(func(ctx context.Context, input *elbv2.DescribeTagsInput, d Deps, accountID string) (any, error) {
+		return DescribeTags(ctx, input, d.NATS, accountID)
+	}),
+	"AddTags": handler(func(ctx context.Context, input *elbv2.AddTagsInput, d Deps, accountID string) (any, error) {
+		return AddTags(ctx, input, d.NATS, accountID)
+	}),
+	"RemoveTags": handler(func(ctx context.Context, input *elbv2.RemoveTagsInput, d Deps, accountID string) (any, error) {
+		return RemoveTags(ctx, input, d.NATS, accountID)
+	}),
+	"LBAgentHeartbeat": handler(func(ctx context.Context, input *elbv2domain.LBAgentHeartbeatInput, d Deps, accountID string) (any, error) {
+		return LBAgentHeartbeat(ctx, input, d.NATS, accountID)
+	}),
+	"GetLBConfig": handler(func(ctx context.Context, input *elbv2domain.GetLBConfigInput, d Deps, accountID string) (any, error) {
+		return GetLBConfig(ctx, input, d.NATS, accountID)
+	}),
+	"ModifyTargetGroup": handler(func(ctx context.Context, input *elbv2.ModifyTargetGroupInput, d Deps, accountID string) (any, error) {
+		return ModifyTargetGroup(ctx, input, d.NATS, accountID)
+	}),
+	"ModifyTargetGroupAttributes": handler(func(ctx context.Context, input *elbv2.ModifyTargetGroupAttributesInput, d Deps, accountID string) (any, error) {
+		return ModifyTargetGroupAttributes(ctx, input, d.NATS, accountID)
+	}),
+	"DescribeTargetGroupAttributes": handler(func(ctx context.Context, input *elbv2.DescribeTargetGroupAttributesInput, d Deps, accountID string) (any, error) {
+		return DescribeTargetGroupAttributes(ctx, input, d.NATS, accountID)
+	}),
+	"ModifyLoadBalancerAttributes": handler(func(ctx context.Context, input *elbv2.ModifyLoadBalancerAttributesInput, d Deps, accountID string) (any, error) {
+		return ModifyLoadBalancerAttributes(ctx, input, d.NATS, accountID)
+	}),
+	"DescribeLoadBalancerAttributes": handler(func(ctx context.Context, input *elbv2.DescribeLoadBalancerAttributesInput, d Deps, accountID string) (any, error) {
+		return DescribeLoadBalancerAttributes(ctx, input, d.NATS, accountID)
+	}),
+	"SetSecurityGroups": handler(func(ctx context.Context, input *elbv2.SetSecurityGroupsInput, d Deps, accountID string) (any, error) {
+		return SetSecurityGroups(ctx, input, d.NATS, accountID)
+	}),
+	"SetIpAddressType": handler(func(ctx context.Context, input *elbv2.SetIpAddressTypeInput, d Deps, accountID string) (any, error) {
+		return SetIpAddressType(ctx, input, d.NATS, accountID)
+	}),
+	"SetSubnets": handler(func(ctx context.Context, input *elbv2.SetSubnetsInput, d Deps, accountID string) (any, error) {
+		return SetSubnets(ctx, input, d.NATS, accountID)
+	}),
+	"AddListenerCertificates": handler(func(ctx context.Context, input *elbv2.AddListenerCertificatesInput, d Deps, accountID string) (any, error) {
+		return AddListenerCertificates(ctx, input, d.NATS, accountID)
+	}),
+	"RemoveListenerCertificates": handler(func(ctx context.Context, input *elbv2.RemoveListenerCertificatesInput, d Deps, accountID string) (any, error) {
+		return RemoveListenerCertificates(ctx, input, d.NATS, accountID)
+	}),
+	"DescribeListenerCertificates": handler(func(ctx context.Context, input *elbv2.DescribeListenerCertificatesInput, d Deps, accountID string) (any, error) {
+		return DescribeListenerCertificates(ctx, input, d.NATS, accountID)
+	}),
+	"DescribeSSLPolicies": handler(func(ctx context.Context, input *elbv2.DescribeSSLPoliciesInput, d Deps, accountID string) (any, error) {
+		return DescribeSSLPolicies(ctx, input, d.NATS, accountID)
+	}),
+	"DescribeAccountLimits": handler(func(ctx context.Context, input *elbv2.DescribeAccountLimitsInput, d Deps, accountID string) (any, error) {
+		loadBalancers := DefaultLoadBalancerLimit
+		if d.Quota != nil {
+			limit, err := d.Quota.LoadBalancerLimit(ctx, accountID)
+			if err != nil {
+				return nil, err
+			}
+			loadBalancers = limit
+		}
+		return DescribeAccountLimits(input, loadBalancers), nil
+	}),
+	"DescribeListenerAttributes": handler(func(ctx context.Context, input *DescribeListenerAttributesInput, d Deps, accountID string) (any, error) {
+		return DescribeListenerAttributes(input, accountID)
+	}),
+	"ModifyListenerAttributes": handler(func(ctx context.Context, input *ModifyListenerAttributesInput, d Deps, accountID string) (any, error) {
+		return ModifyListenerAttributes(input, accountID)
+	}),
+}
+
+// NewRegistration returns the ELBv2 AWS Query control-plane registration.
+func NewRegistration(deps Deps) dispatch.Registration {
+	return dispatch.Registration{
+		Service:       ServiceName,
+		InventoryName: InventoryName,
+		ResolveAction: resolveAction,
+		Dispatch:      deps.dispatch,
+		Errors:        dispatch.ErrorEnvelopeXML,
+		Inventory:     OperationInventory(),
+	}
+}
+
+// ActionNames lists every action the registration dispatches, sorted.
+func ActionNames() []string {
+	return slices.Sorted(maps.Keys(actions))
+}
+
+// OperationInventory classifies the actions the registration serves. Every
+// ELBv2 action is implemented, so none is Stubbed or Unsupported.
+func OperationInventory() dispatch.Inventory {
+	return dispatch.Inventory{Registered: ActionNames()}
+}
+
+// resolveAction is the one Query Action parser telemetry and dispatch share.
+// It leaves the body readable for the dispatcher and answers "" on any failure.
+func resolveAction(r *http.Request) string {
+	args, err := readQueryArgs(r)
+	if err != nil {
+		return ""
+	}
+	return args["Action"]
+}
+
+// readQueryArgs parses the Query body and rewinds it, so the same bytes can be
+// read again. Any read or parse failure is a malformed query string.
+func readQueryArgs(r *http.Request) (map[string]string, error) {
+	body, err := io.ReadAll(io.LimitReader(r.Body, sigv4.MaxPayloadLen+1))
+	if err != nil {
+		return nil, err
+	}
+	r.Body = io.NopCloser(bytes.NewReader(body))
+	if int64(len(body)) > sigv4.MaxPayloadLen {
+		return nil, errors.New(awserrors.ErrorRequestEntityTooLarge)
+	}
+	return query.ParseArgs(string(body))
+}
+
+// dispatch parses, scopes and authorizes the request from one parse of its
+// body, then runs the action. Errors are awserrors codes the gateway renders.
+func (d Deps) dispatch(w http.ResponseWriter, inv dispatch.Invocation) error {
+	r := inv.Request
+	queryArgs, err := readQueryArgs(r)
+	if err != nil {
+		slog.DebugContext(r.Context(), "ELBv2: malformed query string", "err", err)
+		return errors.New(awserrors.ErrorMalformedQueryString)
+	}
+
+	name := queryArgs["Action"]
+	if name == "" {
+		return errors.New(awserrors.ErrorMissingAction)
+	}
+	h, ok := actions[name]
+	if !ok {
+		return errors.New(awserrors.ErrorInvalidAction)
+	}
+
+	// inv.Region is the gateway's own region, never the caller-supplied
+	// credential-scope region, which would let a caller sign for another region
+	// and slide out from under a Deny.
+	accountID := inv.AccountID
+	if accountID == "" {
+		slog.ErrorContext(r.Context(), "ELBv2 dispatch: no account ID in auth context")
+		// InternalError, not ServerInternal: the policy gate used to reach this
+		// case first and that is the code the caller has always seen.
+		return errors.New(awserrors.ErrorInternalError)
+	}
+
+	input, err := h.parse(queryArgs)
+	if err != nil {
+		if errors.Is(err, query.ErrSliceTooLarge) {
+			return errors.New(awserrors.ErrorMalformedQueryString)
+		}
+		return err
+	}
+
+	resources, err := ResourceARNs(name, inv.Region, accountID, input)
+	if err != nil {
+		return err
+	}
+	if err := inv.Authorize(ServiceName, name, resources, nil); err != nil {
+		return err
+	}
+
+	if d.NATS == nil {
+		return errors.New(awserrors.ErrorServerInternal)
+	}
+
+	xmlOutput, err := h.dispatch(r.Context(), name, input, d, accountID)
+	if err != nil {
+		return err
+	}
+
+	w.Header().Set("Content-Type", "text/xml")
+	w.WriteHeader(http.StatusOK)
+	if _, err := w.Write(xmlOutput); err != nil {
+		slog.ErrorContext(r.Context(), "Failed to write ELBv2 response", "err", err)
+	}
+	return nil
+}
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
joshsiv-mulga	6bd7028329	docs: point the /admin migration row at spinifex#1345	2026-10-30 10:50:52 +10:00
joshsiv-mulga	237fcdff94	docs: record the /admin move into the spinifex registration	2026-10-30 10:24:04 +10:00
joshsiv-mulga	5dd95b1ecd	refactor: serve /admin through the spinifex registration	2026-10-30 10:23:37 +10:00
joshsiv-mulga	63cf1beaa7	refactor: move the /admin handlers and wire types into operator/admin	2026-10-30 10:07:21 +10:00
Josh Siviour	ae8ee1e5fe	refactor: register RDS through the ingress seam (#1344 ) * refactor(gateway): register RDS through the ingress seam RDS dispatch moves from gateway/rds.go to domains/rds/awsapi.NewRegistration, composed by awsgw and the integration harness. The caller comes from dispatch.Actor and the principal-class gate still runs before the policy check. * docs: point the RDS migration row at spinifex#1344 * chore: group the imports merged in the actor projection test	2026-10-30 09:59:15 +10:00
Josh Siviour	488f1c7bf9	feat: SQS store guest on a prepared cold-attached EBS volume (not wired) (#1343 ) * feat: build a separate SQS store microVM image Package the go.mod-pinned nats-server, ext4 support and a clean ACPI power-down into a microVM kernel and initramfs of their own. The build checks the packed archive's contents and refuses a nats-server whose version differs from the pin. The image does not start the store and nothing installs or releases it; the LB image is unchanged. * fix: publish the SQS store image only after its gates pass Stage the kernel and initramfs and move them into place only once the content, size and guest-RAM gates pass, so a failed build cannot leave a pair make takes as up to date. The old initramfs is removed first and the new one lands last, so an interrupted publish never leaves a mixed pair. Check the module list without piping into grep -q, which under pipefail reported a present module as missing. * feat: start the SQS store guest on its checked data disk The SQS store image now prepares, runs and stops a dedicated JetStream server. It reads a narrow fw_cfg contract (data disk serial, listen address, one user NKey public key), refuses anything absent, malformed or ambiguous, has the pinned server test the config, then checks and mounts exactly one matching ext4 disk. It never formats and has no ephemeral fallback. The server listens only on the supplied guest NIC address, admits only that NKey and syncs every write. Shutdown stops the recorded run process first, whether it is the server or a start not yet exec'd, and runs the generic read-only remount only once it is gone. A store that survives SIGKILL leaves every mount alone. * fix: report the SQS store as prepared, not listening, before it starts * feat: launch the SQS store guest on its own versioned subject An explicit SQS store kind carries an existing system-account data volume and the one NATS user public key the store accepts. It is served only on system.LaunchSQSStore.v1, so a daemon without this code has no responder and the launch fails instead of booting a load balancer without its disk; the legacy launch path refuses the kind on both send and receive. Every refusal happens before capacity, an ENI or a VM record exists. The volume goes through the normal EBS mount and terminate detach and is never deleted on termination. The store boots its own image with CPU mitigations on, gets only its network, disk serial, listen address and key, and cannot be relaunched after a host restart yet. * feat: ship the SQS store microVM image in releases and installs Release builds now build the SQS store image and stage it beside the load balancer's as microvm-sqs-store/, so an older installer that copies microvm/* without -r still works. setup.sh installs it to /usr/share/spinifex/microvm/sqs-store/, where the daemon looks for it. A release without a complete image never leaves an older one usable: staging drops a previously staged image, and the installer removes an earlier installed pair before warning (missing) or failing (incomplete). Only the two image files are removed. The initramfs is installed last, so an interrupted install leaves no bootable mismatched pair. * feat: restart the SQS store guest on its launching node Persist the non-secret guest config of an SQS store instance (launching node, data volume, listen address, client public NKey and rendered network config) in the instance record, and rebuild its fw_cfg files from it on relaunch. The relaunch refuses on any other node, or when the record is missing or malformed, before writing any file; the data volume is remounted through the normal EBS path and kept on terminate. Cross-node recovery now never selects or claims an SQS store record, so no other node spends retry budget on it or relaunches it. Same-node restore is unchanged, as are load balancer restarts. Older daemons know neither the exclusion nor the record field, so this does not prevent a cross-node relaunch by a mixed-version cluster. * feat: hold prepared system volumes in creating until released Add an internal CreateVolumeForPreparation path that writes a new volume's first record as creating, and CompleteVolumePreparation to release an unattached creating volume in the caller's account to available. Ordinary CreateVolume is unchanged. Attach and delete already refuse a creating volume. ModifyVolume, ForceDetachVolume and CreateSnapshot now refuse it with IncorrectState so nothing can expand, release or capture a volume mid-preparation. * feat: prepare fresh SQS store volumes with an ext4 filesystem The SQS store guest never formats a disk, so a new volume gets its filesystem on the host. The preparer checks for mkfs.ext4 and qemu-img, builds an ext4 image in private scratch, creates the volume in "creating", publishes it on this node, copies the image with the NBD import helper, unpublishes, and only then releases it as available. It never writes a volume it did not just create. A volume is deleted only when no export can still be live: after a confirmed unpublish, or when the export was not NBD. Any publish error, a failed or ambiguous unpublish, or a failed release retains the volume and reports its ID. A publish refused as in-use or already-exists sends no unpublish, since another export may exist. Cleanup ignores caller cancellation but is bounded at 30s. The raw-image copy moves to runtime/compute/nbd unchanged, and the domain receives it as an injected writer. The volume service gains a discard path for an unattached volume still in "creating". * fix: report unknown state and match qemu-img lookup in volume prep A failed release may still have made the volume available, so the retained-volume error now says its state is unknown rather than claiming it is still creating. The copy runs qemu-img by name through PATH, so the pre-allocation check now uses a PATH lookup for it too. mkfs.ext4 keeps the sbin fallback because it runs from the resolved path. * docs: describe the unwired SQS store system instance Records what the guest image, versioned launch, same-node restart and fresh-volume preparation provide, that nothing calls them, and what is untested or missing before production use.	2026-10-30 08:16:43 +10:00
Ben Duncan	3617e4a045	fix: log an expected outcome at its own level, not ERROR (#1342 ) A dispatch failure was logged at ERROR whatever the caller was about to be given, so DryRunOperation, which AWS documents as the success answer to a dry run, read as a fault of ours along with every NotFound and malformed parameter. The level now comes from the status the caller receives, resolved the way ErrorHandler resolves it. QEMU announces its own exit on stderr, so a guest stopping on the SIGTERM the daemon sent it was logged at ERROR too.	2026-10-29 14:38:41 -05:00
Josh Siviour	24281cb860	fix: count volume lease validity from the write, not its reply (#1340 ) * fix: count volume lease validity from the write, not its reply The server starts a lease entry's TTL when it applies the write, but the holder started its validity window when the acknowledgement arrived, so a slow reply stretched validity past the server's expiry and a peer could acquire the volume while the holder still considered it valid. Validity now counts from when each create, generation publish and renewal was sent. An acquisition acknowledged after validity lapsed is refused as retryable, and a lost or lapsed lease still awaiting its fence is no longer shared with another open on the same node. A renewal that lands after the lease was given up, or after its validity elapsed, cannot revive it: the lapse check and the update run under the lease lock, and fencing runs once. The renew loop also fences at the validity deadline itself rather than at the next check, which could consume the whole margin. This bounds local validity only; it is not backend-enforced fencing, and a paused host can still write after its validity ends. * docs: describe volume lease validity, fencing and their limits	2026-10-29 21:38:22 +10:00
Josh Siviour	e58838bc67	refactor: verify SigV4 credentials through an ingress capability (#1339 ) * refactor(gateway): verify SigV4 credentials behind a CredentialVerifier The authenticator no longer imports the IAM and STS domains. It hands the signature check to a verifier as a callback, so the secret never leaves the adapter, and runs the account-active check between Authenticate and Resolve. A nil verifier fails every request closed. The adapter lives in runtime/roles/awsgw/credverify. ResolveCallerUser and the principal-type constants move from domains/sts/awsapi to domains/sts so the adapter can use them without importing an awsapi package. * fix(gateway): guard a nil credential verifier and tighten its tests A typed-nil credverify.Verifier now fails closed instead of panicking. The session verdict test goes through the real verifier into the lockout boundary. Test setups build the verifier from the same mock as the gateway services. * test(gateway): commit the verifier test helper	2026-10-29 21:30:06 +10:00
Josh Siviour	f0e4ffe70c	refactor: give seam services the verified caller's identity facts (#1338 ) * test: pin the caller facts the gateway gives a registered service per principal kind * refactor(gateway): seal the verified caller into one request-context value * feat(dispatch): give registered services the caller's principal ARN, user ID, user name and session name Actor gains PrincipalARN, UserID, UserName and SessionName, filled from the sealed principal, and RoleName(), which keeps the account-checked role ARN rule in one place. No consumer reads them yet. * refactor(gateway): read the caller from dispatch.Actor in the RDS, EKS, STS, IAM and admin handlers callerActor builds the Actor the seam hands services, and the seven readers that took fields off the sealed principal now take them from it. RDS and EKS take the role name from Actor.RoleName(). No behaviour change.	2026-10-29 16:13:41 +10:00
Josh Siviour	18adea9486	refactor: register the spinifex operator service through the ingress seam (#1337 ) * test: characterize the spinifex Query and /admin surfaces through the full gateway pipeline * refactor: register the spinifex Query service from operator/admin * docs: point the spinifex registration migration row at the PR	2026-10-29 14:55:50 +10:00
Julian Sommer	f06def6894	chore: update operation count	2026-10-29 13:50:59 +10:00
Julian Sommer	ef7d30578f	chore(awsmodel): sync AWS models to api-models-aws e10d6a1c (2026-10-28) Seven models changed and add 26 unimplemented operations. The request sweep gains 33 requests with no new findings.	2026-10-29 13:35:39 +10:00
Julian Sommer	b76679e571	fix(integration): restore ELBv2 to the request conformance sweep The ELBv2 move to the ingress seam added its inventory to the coverage tool but not to TestRequestConformance, so 36 operations and 45 findings dropped out while the suite passed. A served service with no implemented operations now fails the sweep.	2026-10-29 13:34:12 +10:00
Josh Siviour	d7d922e963	refactor: register Tagging and ELBv2 through the ingress seam (#1336 ) * test: characterize Tagging dispatch, authorization and GetResources responses through the gateway pipeline * refactor: register Resource Groups Tagging through the ingress seam * test: characterize ELBv2 through the full gateway pipeline * refactor: register ELBv2 through the ingress seam * test: cover the ELBv2 registration, query parser and load balancer limit * test: characterize RDS through the full gateway pipeline * docs: point the Tagging and ELBv2 migration rows at the PR * test: register Tagging and ELBv2 on the integration harness gateway	2026-10-29 13:27:43 +10:00
Julian Sommer	64e3711893	fix(awsmodel): accept InvalidGatewayID.NotFound in the EC2 error catalog AWS's EC2 error reference documents it, and DeleteEgressOnlyInternetGateway and ReplaceRoute answer it, so the integration suite reported two false violations.	2026-10-29 13:27:09 +10:00
Josh Siviour	ad9568a6c6	refactor: move gateway/spx, admin wire types and EKS OIDC discovery out of gateway (#1334 ) * refactor: move gateway/spx to operator/admin * test: pin the admin wire types and method list in golden JSON * refactor: move the admin wire types and method list to contracts/operator/v1 * refactor: serve EKS OIDC discovery and JWKS from the EKS domain * refactor: assert the EKS OIDC publisher satisfies the gateway interface * docs: cite the pull request in the gateway root-move rows	2026-10-29 12:22:16 +10:00
Josh Siviour	dedb3d8d75	ci: keep the E2E WAN packet captures only for runs that did not pass (#1335 ) The captures are about 90% of each E2E artifact and nothing reads them unless a run failed, so a clean single-node or multi-node pass drops them before the artifact upload.	2026-10-29 12:21:30 +10:00
Julian Sommer	cc1186fa48	fix: grant the build user passwordless sudo for the appliance build when it lacks it, and remove the grant at seal	2026-10-29 11:10:51 +10:00
Julian Sommer	dbdb827a69	fix: install tar on a minimal Oracle Linux build system before copying the appliance payload	2026-10-29 11:05:35 +10:00
Josh Siviour	c94b0f22f6	feat: accept EnaSupport on RegisterImage, report it on instances, and name refused parameters (#1332 ) * feat: accept EnaSupport on RegisterImage and name the refused parameter in its errors EnaSupport true and false are stored on the image record, returned by DescribeImages when set and carried through CopyImage. Every validation error in ValidateRegisterImageInput keeps its code and now says which parameter and value it refused. * feat: set EnaSupport on images created from an instance AWS takes an image's EnaSupport from the instance it was made from, and every instance here reports it, so the source AMI's value is not inherited. * docs: record RegisterImage EnaSupport support and the refused parameters * fix: report EnaSupport on DescribeInstances and RunInstances	2026-10-29 10:53:30 +10:00
Julian Sommer	71567d0998	fix: show the appliance password rules before the first-login change	2026-10-29 10:51:06 +10:00
Josh Siviour	11e8b68ddc	refactor: remove the unused host SQS message store service (#1333 ) * refactor: remove the host SQS store systemd unit The SQS message store will run inside system-owned instances rather than as a host daemon, so spinifex.target no longer starts it. Bump the target to version 4 and drop the unit from the embedded set and its invariants. * refactor: remove host SQS store config provisioning The SQS store will run in system-owned instances, so admin init and join no longer generate or distribute a host store token, routes or config, and the formation join payload drops the store token field. * refactor: remove host SQS store service command and role The SQS store will run in system-owned instances, so drop the unused spx service sqs-store command, its service registration and the host sqsstore role. The domains/sqs/message JetStream store is unaffected. * refactor: remove host SQS store provisioning from install scripts The SQS store will run in system-owned instances, so setup no longer creates the spinifex-sqsstore user, its directories or the 4223/4249 peer firewall ports, and install, reset and uninstall stop handling the retired unit, user and data directory. * docs: drop the host SQS store from firewall and connection tables Spinifex no longer ships a host SQS store service, so remove its ports from the host firewall and network connection references and say in DESIGN.md that no host service, unit, user or firewall rule exists for it.	2026-10-29 10:48:05 +10:00
Julian Sommer	8f4754da13	fix: install the appliance with the setup.sh from the release tarball, since install.mulgadc.com serves the stable one	2026-10-29 10:30:48 +10:00
Julian Sommer	8b965d24ce	fix: refuse Route 53 requests that break the model's bounds with InvalidInput (#1329 )	2026-10-29 11:29:45 +11:00
Julian Sommer	ac2620df58	fix: report an SSH failure in build-oracle-appliance.sh as one, not as missing sudo	2026-10-29 10:26:09 +10:00
Julian Sommer	2feb264863	feat: Oracle Linux multi-node clusters with OVN IPsec (#1327 )	2026-10-29 11:17:22 +11:00
Josh Siviour	ddadfa2812	chore: remove unused e2e suite-selection manifest and drift lints (#1331 ) The targeted suite selector (select-suites) has not run in CI since its shadow-mode step was removed in #556. The service-interfaces manifest, manifest-check and manifest-lint existed only to feed it, so they go too. Also drops the now-meaningless e2e:allow-create markers, keeping their rationale as plain comments.	2026-10-29 10:04:31 +10:00
Josh Siviour	5923143062	feat: set X-Forwarded-For, -Proto and -Port on ALB requests (#1330 ) * feat: render X-Forwarded-For, -Proto and -Port headers in ALB HAProxy frontends Append joins every client X-Forwarded-For line and adds the client address as AWS does, preserve and remove follow the load balancer attribute, and Proto and Port are always overwritten. * feat: validate the ALB forwarding attributes and re-render config when they change ModifyLoadBalancerAttributes rejects bad xff_header_processing.mode and xff_client_port.enabled values with ValidationError before storing anything, and regenerates the stored config when either key changes. * test: assert the ALB X-Forwarded headers and the remove mode in the lb e2e The app instances echo the request headers on /__headers. The internal ALB is driven from the probe client and the HTTPS ALB from the runner, covering the caller address, a client value, spoofed Proto and Port, mode remove and a rejected mode. * test: let the lb probe client trigger accept a null header list * docs: document the ALB forwarding headers and their attributes	2026-10-29 09:33:39 +10:00
Julian Sommer	46ca3dfba1	chore: comment	2026-10-29 09:13:32 +10:00