bump version number to v2.3.0

db: avoid iterating bounded from empty slice (fix #1401 ) (#1408 )
Reviewed-on: https://git.deuxfleurs.fr/Deuxfleurs/garage/pulls/1408 Co-authored-by: Alex Auvolat <lx@deuxfleurs.fr> Co-committed-by: Alex Auvolat <lx@deuxfleurs.fr>
2026-08-13 15:46:53 +00:00 · 2026-04-16 18:34:27 +02:00 · 2026-04-16 16:33:28 +00:00 · 2026-04-16 16:22:17 +00:00 · 2026-04-16 16:22:17 +00:00 · 2026-04-16 16:20:27 +00:00
487 changed files with 84526 additions and 22924 deletions
@@ -1,288 +0,0 @@
---
-kind: pipeline
-name: default
-
-node:
-  nix-daemon: 1
-
-steps:
-  - name: check formatting
-    image: nixpkgs/nix:nixos-22.05
-    commands:
-      - nix-shell --attr rust --run "cargo fmt -- --check"
-
-  - name: build
-    image: nixpkgs/nix:nixos-22.05
-    commands:
-      - nix-build --no-build-output --attr clippy.amd64 --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
-
-  - name: unit + func tests
-    image: nixpkgs/nix:nixos-22.05
-    environment:
-      GARAGE_TEST_INTEGRATION_EXE: result-bin/bin/garage
-    commands:
-      - nix-build --no-build-output --attr clippy.amd64 --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-build --no-build-output --attr test.amd64
-      - ./result/bin/garage_db-*
-      - ./result/bin/garage_api-*
-      - ./result/bin/garage_model-*
-      - ./result/bin/garage_rpc-*
-      - ./result/bin/garage_table-*
-      - ./result/bin/garage_util-*
-      - ./result/bin/garage_web-*
-      - ./result/bin/garage-*
-      - ./result/bin/integration-*
-      - rm result
-
-  - name: integration tests
-    image: nixpkgs/nix:nixos-22.05
-    commands:
-      - nix-build --no-build-output --attr clippy.amd64 --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --attr integration --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
-
-trigger:
-  event:
-  - custom
-  - push
-  - pull_request
-  - tag
-  - cron
-
---
-kind: pipeline
-type: docker
-name: release-linux-amd64
-
-node:
-  nix-daemon: 1
-
-steps:
-  - name: build
-    image: nixpkgs/nix:nixos-22.05
-    commands:
-      - nix-build --no-build-output --attr pkgs.amd64.release --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --attr rust --run "./script/not-dynamic.sh result-bin/bin/garage"
-
-  - name: integration
-    image: nixpkgs/nix:nixos-22.05
-    commands:
-      - nix-shell --attr integration --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
-
-  - name: push static binary
-    image: nixpkgs/nix:nixos-22.05
-    environment:
-      AWS_ACCESS_KEY_ID:
-        from_secret: garagehq_aws_access_key_id
-      AWS_SECRET_ACCESS_KEY:
-        from_secret: garagehq_aws_secret_access_key
-      TARGET: "x86_64-unknown-linux-musl"
-    commands:
-      - nix-shell --attr release --run "to_s3"
-
-  - name: docker build and publish
-    image: nixpkgs/nix:nixos-22.05
-    environment:
-      DOCKER_AUTH:
-        from_secret: docker_auth
-      DOCKER_PLATFORM: "linux/amd64"
-      CONTAINER_NAME: "dxflrs/amd64_garage"
-      HOME: "/kaniko"
-    commands:
-      - mkdir -p /kaniko/.docker
-      - echo $DOCKER_AUTH > /kaniko/.docker/config.json
-      - export CONTAINER_TAG=${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --attr release --run "to_docker"
-
-
-trigger:
-  event:
-  - promote
-  - cron
-
---
-kind: pipeline
-type: docker
-name: release-linux-i386
-
-node:
-  nix-daemon: 1
-
-steps:
-  - name: build
-    image: nixpkgs/nix:nixos-22.05
-    commands:
-      - nix-build --no-build-output --attr pkgs.i386.release --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --attr rust --run "./script/not-dynamic.sh result-bin/bin/garage"
-
-  - name: integration
-    image: nixpkgs/nix:nixos-22.05
-    commands:
-      - nix-shell --attr integration --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
-
-  - name: push static binary
-    image: nixpkgs/nix:nixos-22.05
-    environment:
-      AWS_ACCESS_KEY_ID:
-        from_secret: garagehq_aws_access_key_id
-      AWS_SECRET_ACCESS_KEY:
-        from_secret: garagehq_aws_secret_access_key
-      TARGET: "i686-unknown-linux-musl"
-    commands:
-      - nix-shell --attr release --run "to_s3"
-
-  - name: docker build and publish
-    image: nixpkgs/nix:nixos-22.05
-    environment:
-      DOCKER_AUTH:
-        from_secret: docker_auth
-      DOCKER_PLATFORM: "linux/386"
-      CONTAINER_NAME: "dxflrs/386_garage"
-      HOME: "/kaniko"
-    commands:
-      - mkdir -p /kaniko/.docker
-      - echo $DOCKER_AUTH > /kaniko/.docker/config.json
-      - export CONTAINER_TAG=${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --attr release --run "to_docker"
-
-trigger:
-  event:
-  - promote
-  - cron
-
---
-kind: pipeline
-type: docker
-name: release-linux-arm64
-
-node:
-  nix-daemon: 1
-
-steps:
-  - name: build
-    image: nixpkgs/nix:nixos-22.05
-    commands:
-      - nix-build --no-build-output --attr pkgs.arm64.release --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --attr rust --run "./script/not-dynamic.sh result-bin/bin/garage"
-
-  - name: push static binary
-    image: nixpkgs/nix:nixos-22.05
-    environment:
-      AWS_ACCESS_KEY_ID:
-        from_secret: garagehq_aws_access_key_id
-      AWS_SECRET_ACCESS_KEY:
-        from_secret: garagehq_aws_secret_access_key
-      TARGET: "aarch64-unknown-linux-musl"
-    commands:
-      - nix-shell --attr release --run "to_s3"
-
-  - name: docker build and publish
-    image: nixpkgs/nix:nixos-22.05
-    environment:
-      DOCKER_AUTH:
-        from_secret: docker_auth
-      DOCKER_PLATFORM: "linux/arm64"
-      CONTAINER_NAME: "dxflrs/arm64_garage"
-      HOME: "/kaniko"
-    commands:
-      - mkdir -p /kaniko/.docker
-      - echo $DOCKER_AUTH > /kaniko/.docker/config.json
-      - export CONTAINER_TAG=${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --attr release --run "to_docker"
-
-trigger:
-  event:
-  - promote
-  - cron
-
---
-kind: pipeline
-type: docker
-name: release-linux-arm
-
-node:
-  nix-daemon: 1
-
-steps:
-  - name: build
-    image: nixpkgs/nix:nixos-22.05
-    commands:
-      - nix-build --no-build-output --attr pkgs.arm.release --argstr git_version ${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --attr rust --run "./script/not-dynamic.sh result-bin/bin/garage"
-
-  - name: push static binary
-    image: nixpkgs/nix:nixos-22.05
-    environment:
-      AWS_ACCESS_KEY_ID:
-        from_secret: garagehq_aws_access_key_id
-      AWS_SECRET_ACCESS_KEY:
-        from_secret: garagehq_aws_secret_access_key
-      TARGET: "armv6l-unknown-linux-musleabihf"
-    commands:
-      - nix-shell --attr release --run "to_s3"
-
-  - name: docker build and publish
-    image: nixpkgs/nix:nixos-22.05
-    environment:
-      DOCKER_AUTH:
-        from_secret: docker_auth
-      DOCKER_PLATFORM: "linux/arm"
-      CONTAINER_NAME: "dxflrs/arm_garage"
-      HOME: "/kaniko"
-    commands:
-      - mkdir -p /kaniko/.docker
-      - echo $DOCKER_AUTH > /kaniko/.docker/config.json
-      - export CONTAINER_TAG=${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --attr release --run "to_docker"
-
-trigger:
-  event:
-  - promote
-  - cron
-
---
-kind: pipeline
-type: docker
-name: refresh-release-page
-
-node:
-  nix-daemon: 1
-
-steps:
-  - name: multiarch-docker
-    image: nixpkgs/nix:nixos-22.05
-    environment:
-      DOCKER_AUTH:
-        from_secret: docker_auth
-      HOME: "/root"
-    commands:
-      - mkdir -p /root/.docker
-      - echo $DOCKER_AUTH > /root/.docker/config.json
-      - export CONTAINER_TAG=${DRONE_TAG:-$DRONE_COMMIT}
-      - nix-shell --attr release --run "multiarch_docker"
-  - name: refresh-index
-    image: nixpkgs/nix:nixos-22.05
-    environment:
-      AWS_ACCESS_KEY_ID:
-        from_secret: garagehq_aws_access_key_id
-      AWS_SECRET_ACCESS_KEY:
-        from_secret: garagehq_aws_secret_access_key
-    commands:
-      - mkdir -p /etc/nix && cp nix/nix.conf /etc/nix/nix.conf
-      - nix-shell --attr release --run "refresh_index"
-
-depends_on:
-  - release-linux-amd64
-  - release-linux-i386
-  - release-linux-arm64
-  - release-linux-arm
-
-trigger:
-  event:
-  - promote
-  - cron
-
---
-kind: signature
-hmac: ac09a5a8c82502f67271f93afa1e1e21ce66383b8e24a6deb26b285cc1c378ba
-
-...
@@ -0,0 +1,55 @@
+labels:
+  nix: "enabled"
+
+when:
+  - event:
+    - tag
+    - pull_request
+    - deployment
+    - cron
+    - manual
+  - event: push
+    branch: main-*
+
+steps:
+  - name: check formatting
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-build -j4 --attr flakePackages.fmt
+
+  - name: check typos
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-shell --attr ci --run typos
+
+  - name: check lints with clippy
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-build -j4 --attr flakePackages.clippy
+
+  - name: build
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-build -j4 --attr flakePackages.dev
+
+  - name: unit + func tests (lmdb)
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-build -j4 --attr flakePackages.tests-lmdb
+
+  - name: unit + func tests (sqlite)
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-build -j4 --attr flakePackages.tests-sqlite
+
+  - name: unit + func tests (fjall)
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-build -j4 --attr flakePackages.tests-fjall
+
+  - name: integration tests
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-build -j4 --attr flakePackages.dev
+      - nix-shell --attr ci --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
+    depends_on: [ build ]
@@ -0,0 +1,33 @@
+labels:
+  nix: "enabled"
+
+when:
+  event:
+  - deployment
+  - cron
+
+depends_on:
+  - release
+
+steps:
+  - name: refresh-index
+    image: nixpkgs/nix:nixos-24.05
+    environment:
+      AWS_ACCESS_KEY_ID:
+        from_secret: garagehq_aws_access_key_id
+      AWS_SECRET_ACCESS_KEY:
+        from_secret: garagehq_aws_secret_access_key
+    commands:
+      - mkdir -p /etc/nix && cp nix/nix.conf /etc/nix/nix.conf
+      - nix-shell --attr ci --run "refresh_index"
+
+  - name: multiarch-docker
+    image: nixpkgs/nix:nixos-24.05
+    environment:
+      DOCKER_AUTH:
+        from_secret: docker_auth
+    commands:
+      - mkdir -p /root/.docker
+      - echo $DOCKER_AUTH > /root/.docker/config.json
+      - export CONTAINER_TAG=${CI_COMMIT_TAG:-$CI_COMMIT_SHA}
+      - nix-shell --attr ci --run "multiarch_docker"
@@ -0,0 +1,79 @@
+labels:
+  nix: "enabled"
+
+when:
+  event:
+  - deployment
+  - cron
+
+matrix:
+  include:
+    - ARCH: amd64
+      TARGET: x86_64-unknown-linux-musl
+    - ARCH: i386
+      TARGET: i686-unknown-linux-musl
+    - ARCH: arm64
+      TARGET: aarch64-unknown-linux-musl
+    - ARCH: arm
+      TARGET: armv6l-unknown-linux-musleabihf
+
+steps:
+  - name: build
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-build --attr releasePackages.${ARCH} --argstr git_version ${CI_COMMIT_TAG:-$CI_COMMIT_SHA}
+
+  - name: check is static binary
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-shell --attr ci --run "./script/not-dynamic.sh result/bin/garage"
+
+  - name: integration tests
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-shell --attr ci --run ./script/test-smoke.sh || (cat /tmp/garage.log; false)
+    when:
+      - matrix:
+          ARCH: amd64
+      - matrix:
+          ARCH: i386
+
+  - name: upgrade tests from v1.0.0
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-shell --attr ci --run "./script/test-upgrade.sh v1.0.0 x86_64-unknown-linux-musl" || (cat /tmp/garage.log; false)
+    when:
+      - matrix:
+          ARCH: amd64
+
+  - name: upgrade tests from v0.8.4
+    image: nixpkgs/nix:nixos-24.05
+    commands:
+      - nix-shell --attr ci --run "./script/test-upgrade.sh v0.8.4 x86_64-unknown-linux-musl" || (cat /tmp/garage.log; false)
+    when:
+      - matrix:
+          ARCH: amd64
+
+  - name: push static binary
+    image: nixpkgs/nix:nixos-24.05
+    environment:
+      TARGET: "${TARGET}"
+      AWS_ACCESS_KEY_ID:
+        from_secret: garagehq_aws_access_key_id
+      AWS_SECRET_ACCESS_KEY:
+        from_secret: garagehq_aws_secret_access_key
+    commands:
+      - nix-shell --attr ci --run "to_s3"
+
+  - name: docker build and publish
+    image: nixpkgs/nix:nixos-24.05
+    environment:
+      DOCKER_PLATFORM: "linux/${ARCH}"
+      CONTAINER_NAME: "dxflrs/${ARCH}_garage"
+      DOCKER_AUTH:
+        from_secret: docker_auth
+    commands:
+      - mkdir -p /root/.docker
+      - echo $DOCKER_AUTH > /root/.docker/config.json
+      - export CONTAINER_TAG=${CI_COMMIT_TAG:-$CI_COMMIT_SHA}
+      - nix-shell --attr ci --run "to_docker"
@@ -3,11 +3,15 @@ resolver = "2"
 members = [
 	"src/db",
 	"src/util",
+	"src/net",
 	"src/rpc",
 	"src/table",
 	"src/block",
 	"src/model",
-	"src/api",
+	"src/api/common",
+	"src/api/s3",
+	"src/api/k2v",
+	"src/api/admin",
 	"src/web",
 	"src/garage",
 	"src/k2v-client",
@@ -17,19 +21,174 @@ members = [
 default-members = ["src/garage"]

 [workspace.dependencies]
+
+# Internal Garage crates
 format_table = { version = "0.1.1", path = "src/format-table" }
-garage_api = { version = "0.8.5", path = "src/api" }
-garage_block = { version = "0.8.5", path = "src/block" }
-garage_db = { version = "0.8.5", path = "src/db", default-features = false }
-garage_model = { version = "0.8.5", path = "src/model", default-features = false }
-garage_rpc = { version = "0.8.5", path = "src/rpc" }
-garage_table = { version = "0.8.5", path = "src/table" }
-garage_util = { version = "0.8.5", path = "src/util" }
-garage_web = { version = "0.8.5", path = "src/web" }
+garage_api_common = { version = "2.3.0", path = "src/api/common" }
+garage_api_admin = { version = "2.3.0", path = "src/api/admin" }
+garage_api_s3 = { version = "2.3.0", path = "src/api/s3" }
+garage_api_k2v = { version = "2.3.0", path = "src/api/k2v" }
+garage_block = { version = "2.3.0", path = "src/block" }
+garage_db = { version = "2.3.0", path = "src/db", default-features = false }
+garage_model = { version = "2.3.0", path = "src/model", default-features = false }
+garage_net = { version = "2.3.0", path = "src/net" }
+garage_rpc = { version = "2.3.0", path = "src/rpc" }
+garage_table = { version = "2.3.0", path = "src/table" }
+garage_util = { version = "2.3.0", path = "src/util" }
+garage_web = { version = "2.3.0", path = "src/web" }
 k2v-client = { version = "0.0.4", path = "src/k2v-client" }

-[profile.dev]
-lto = "off"
+# External crates from crates.io
+arc-swap = "1.8"
+argon2 = "0.5"
+async-trait = "0.1"
+backtrace = "0.3"
+base64 = "0.22"
+blake2 = "0.10"
+bytes = "1.11"
+bytesize = "2.3"
+cfg-if = "1.0"
+chrono = { version = "0.4", features = ["serde"] }
+crc-fast = "1.9"
+crypto-common = "0.1"
+gethostname = "1.1"
+git-version = "0.3"
+hex = "0.4"
+hexdump = "0.1"
+hmac = "0.12"
+itertools = "0.14"
+ipnet = "2.11"
+lazy_static = "1.5"
+md-5 = "0.10"
+mktemp = "0.5"
+nix = { version = "0.31", default-features = false, features = ["fs"] }
+nom = "8.0"
+parking_lot = "0.12"
+parse_duration = "2.1"
+paste = "1.0"
+pin-project = "1.1"
+pnet_datalink = "0.35"
+rand = "0.9"
+sha1 = "0.10"
+sha2 = "0.10"
+timeago = { version = "0.5", default-features = false }
+xxhash-rust = { version = "0.8", default-features = false, features = ["xxh3"] }
+
+aes-gcm = { version = "0.10", features = ["aes", "stream"] }
+sodiumoxide = { version = "0.2.5-0", package = "kuska-sodiumoxide" }
+kuska-handshake = { version = "0.2.0", features = ["default", "async_std"] }
+
+clap = { version = "4.5", features = ["derive", "env"] }
+pretty_env_logger = "0.5"
+structopt = { version = "0.3", default-features = false }
+syslog-tracing = "0.3"
+tracing = "0.1"
+tracing-journald = "0.3"
+tracing-subscriber = { version = "0.3", features = ["env-filter"] }
+
+heed = { version = "0.22", default-features = false, features = [] }
+rusqlite = { version = "0.38", features = ["fallible_uint"] }
+r2d2 = "0.8"
+r2d2_sqlite = "0.32"
+fjall = "2.11"
+
+async-compression = { version = "0.4", features = ["tokio", "zstd"] }
+zstd = { version = "0.13", default-features = false }
+
+quick-xml = { version = "0.39", features = ["serialize"] }
+rmp-serde = "1.3"
+serde = { version = "1.0", default-features = false, features = ["derive", "rc"] }
+serde_bytes = "0.11"
+serde_json = "1.0"
+toml = { version = "0.9", default-features = false, features = ["parse", "serde"] }
+utoipa = { version = "5.4", features = ["chrono"] }
+
+# newer version requires rust edition 2021
+k8s-openapi = { version = "0.27", features = ["v1_35"] }
+kube = { version = "3.0", default-features = false, features = [
+    "runtime",
+    "derive",
+    "client",
+    "rustls-tls",
+] }
+schemars = "1.2"
+reqwest = { version = "0.13", default-features = false, features = [
+    "rustls",
+    "json",
+] }
+
+form_urlencoded = "1.2"
+http = "1.4"
+httpdate = "1.0"
+http-range = "0.1"
+http-body-util = "0.1"
+hyper = { version = "1.8", default-features = false }
+hyper-util = { version = "0.1", features = ["full"] }
+multer = "3.1"
+percent-encoding = "2.3"
+roxmltree = "0.21"
+url = "2.5"
+
+futures = "0.3"
+futures-util = "0.3"
+tokio = { version = "1.49", default-features = false, features = [
+    "rt",
+    "rt-multi-thread",
+    "io-util",
+    "net",
+    "time",
+    "macros",
+    "sync",
+    "signal",
+    "fs",
+] }
+tokio-util = { version = "0.7", features = ["compat", "io"] }
+tokio-stream = { version = "0.1", features = ["net"] }
+socket2 = { version = "0.6", features = ["all"] }
+
+opentelemetry = { version = "0.17", features = ["rt-tokio", "metrics", "trace"] }
+opentelemetry-prometheus = "0.10"
+opentelemetry-otlp = "0.10"
+opentelemetry-contrib = "0.9"
+prometheus = "0.13"
+
+# used by the k2v-client crate only
+aws-sigv4 = { version = "1.3", default-features = false }
+hyper-rustls = { version = "0.27", default-features = false, features = [
+    "http1",
+    "http2",
+    "ring",
+    "rustls-native-certs",
+] }
+log = "0.4"
+thiserror = "2.0"
+
+# ---- used only as build / dev dependencies ----
+assert-json-diff = "2.0"
+rustc_version = "0.4"
+static_init = "1.0"
+aws-smithy-runtime = { version = "1.9", default-features = false, features = [
+    "tls-rustls",
+] }
+aws-sdk-config = { version = "1.99", default-features = false }
+aws-sdk-s3 = { version = "1.121", default-features = false, features = [
+    "rt-tokio",
+] }

 [profile.release]
-debug = true
+lto = "thin"
+codegen-units = 16
+opt-level = 3
+strip = "debuginfo"
+
+[workspace.lints.clippy]
+# pedantic lints configuration
+doc_markdown = "warn"
+format_collect = "warn"
+manual_midpoint = "warn"
+semicolon_if_nothing_returned = "warn"
+unnecessary_semicolon = "warn"
+unnecessary_wraps = "warn"
+
+# nursery lints configuration
+# or_fun_call = "warn" # enable it to help detect non trivial code used in `_or` method
@@ -3,5 +3,5 @@ FROM scratch
 ENV RUST_BACKTRACE=1
 ENV RUST_LOG=garage=info

-COPY result-bin/bin/garage /
+COPY result/bin/garage /
 CMD [ "/garage", "server"]
@@ -1,13 +1,8 @@
-.PHONY: doc all release shell run1 run2 run3
+.PHONY: doc all run1 run2 run3

 all:
-	clear; cargo build
-
-release:
-	nix-build --attr pkgs.amd64.release --no-build-output
-
-shell:
-	nix-shell
+	clear
+	cargo build

 # ----

@@ -1,4 +1,4 @@
-Garage [![Build Status](https://drone.deuxfleurs.fr/api/badges/Deuxfleurs/garage/status.svg?ref=refs/heads/main)](https://drone.deuxfleurs.fr/Deuxfleurs/garage)
+Garage [![status-badge](https://woodpecker.deuxfleurs.fr/api/badges/1/status.svg)](https://woodpecker.deuxfleurs.fr/repos/1)
 ===

 <p align="center" style="text-align:center;">
@@ -3,54 +3,22 @@
 with import ./nix/common.nix;

 let
-  pkgs = import pkgsSrc { };
+  pkgs = import nixpkgs { };
  compile = import ./nix/compile.nix;

-  build_debug_and_release = (target: {
-    debug = (compile {
-      inherit system target git_version pkgsSrc cargo2nixOverlay;
-      release = false;
-    }).workspace.garage { compileMode = "build"; };
-
-    release = (compile {
-      inherit system target git_version pkgsSrc cargo2nixOverlay;
-      release = true;
-    }).workspace.garage { compileMode = "build"; };
-  });
-
-  test = (rustPkgs:
-    pkgs.symlinkJoin {
-      name = "garage-tests";
-      paths =
-        builtins.map (key: rustPkgs.workspace.${key} { compileMode = "test"; })
-        (builtins.attrNames rustPkgs.workspace);
-    });
+  build_release = target: (compile {
+    inherit target system git_version nixpkgs;
+    crane = flake.inputs.crane;
+    rust-overlay = flake.inputs.rust-overlay;
+    release = true;
+  }).garage;

 in {
-  pkgs = {
-    amd64 = build_debug_and_release "x86_64-unknown-linux-musl";
-    i386 = build_debug_and_release "i686-unknown-linux-musl";
-    arm64 = build_debug_and_release "aarch64-unknown-linux-musl";
-    arm = build_debug_and_release "armv6l-unknown-linux-musleabihf";
-  };
-  test = {
-    amd64 = test (compile {
-      inherit system git_version pkgsSrc cargo2nixOverlay;
-      target = "x86_64-unknown-linux-musl";
-      features = [
-        "garage/bundled-libs"
-        "garage/k2v"
-        "garage/sled"
-        "garage/lmdb"
-        "garage/sqlite"
-      ];
-    });
-  };
-  clippy = {
-    amd64 = (compile {
-      inherit system git_version pkgsSrc cargo2nixOverlay;
-      target = "x86_64-unknown-linux-musl";
-      compiler = "clippy";
-    }).workspace.garage { compileMode = "build"; };
+  releasePackages = {
+    amd64 = build_release "x86_64-unknown-linux-musl";
+    i386 = build_release "i686-unknown-linux-musl";
+    arm64 = build_release "aarch64-unknown-linux-musl";
+    arm = build_release "armv6l-unknown-linux-musleabihf";
  };
+  flakePackages = flake.packages.${system};
 }
@@ -1,7 +1,7 @@
 <!DOCTYPE html>
 <html>
  <head>
-    <title>Garage Adminstration API v0</title>
+    <title>Garage administration API v0</title>
    <!-- needed for adaptive design -->
    <meta charset="utf-8"/>
    <meta name="viewport" content="width=device-width, initial-scale=1">
@@ -3,10 +3,10 @@ info:
  version: v0.8.0
  title: Garage Administration API v0+garage-v0.8.0
  description: |
-    Administrate your Garage cluster programatically, including status, layout, keys, buckets, and maintainance tasks.
-    
-    *Disclaimer: The API is not stable yet, hence its v0 tag. The API can change at any time, and changes can include breaking backward compatibility. Read the changelog and upgrade your scripts before upgrading. Additionnaly, this specification is very early stage and can contain bugs, especially on error return codes/types that are not tested yet. Do not expect a well finished and polished product!*
-paths: 
+    Administrate your Garage cluster programmatically, including status, layout, keys, buckets, and maintenance tasks.
+
+    *Disclaimer: The API is not stable yet, hence its v0 tag. The API can change at any time, and changes can include breaking backward compatibility. Read the changelog and upgrade your scripts before upgrading. Additionally, this specification is very early stage and can contain bugs, especially on error return codes/types that are not tested yet. Do not expect a well finished and polished product!*
+paths:
  /status:
    get:
      tags: 
@@ -0,0 +1,24 @@
+<!DOCTYPE html>
+<html>
+  <head>
+    <title>Garage administration API v1</title>
+    <!-- needed for adaptive design -->
+    <meta charset="utf-8"/>
+    <meta name="viewport" content="width=device-width, initial-scale=1">
+    <link href="./css/redoc.css" rel="stylesheet">
+
+    <!--
+    Redoc doesn't change outer page styles
+    -->
+    <style>
+      body {
+        margin: 0;
+        padding: 0;
+      }
+    </style>
+  </head>
+  <body>
+    <redoc spec-url='./garage-admin-v1.yml'></redoc>
+    <script src="./redoc.standalone.js"> </script>
+  </body>
+</html>
@@ -0,0 +1,24 @@
+<!DOCTYPE html>
+<html>
+  <head>
+    <title>Garage administration API v2</title>
+    <!-- needed for adaptive design -->
+    <meta charset="utf-8"/>
+    <meta name="viewport" content="width=device-width, initial-scale=1">
+    <link href="./css/redoc.css" rel="stylesheet">
+
+    <!--
+    Redoc doesn't change outer page styles
+    -->
+    <style>
+      body {
+        margin: 0;
+        padding: 0;
+      }
+    </style>
+  </head>
+  <body>
+    <redoc spec-url='./garage-admin-v2.json'></redoc>
+    <script src="./redoc.standalone.js"> </script>
+  </body>
+</html>
@@ -51,4 +51,4 @@ We are currently building this SDK for [Python](@/documentation/build/python.md#

 More information:
  - [In the reference manual](@/documentation/reference-manual/admin-api.md)
-  - [Full specifiction](https://garagehq.deuxfleurs.fr/api/garage-admin-v0.html)
+  - [Full specification](https://garagehq.deuxfleurs.fr/api/garage-admin-v0.html)
@@ -37,30 +37,84 @@ import (
    "context"
    "fmt"
    "os"
+    "strings"
    garage "git.deuxfleurs.fr/garage-sdk/garage-admin-sdk-golang"
 )

 func main() {
-    // Set Host and other parameters
+    // Initialization
    configuration := garage.NewConfiguration()
    configuration.Host = "127.0.0.1:3903"
-
-
-    // We can now generate a client
    client := garage.NewAPIClient(configuration)
-
-    // Authentication is handled through the context pattern
    ctx := context.WithValue(context.Background(), garage.ContextAccessToken, "s3cr3t")

-    // Send a request
-    resp, r, err := client.NodesApi.GetNodes(ctx).Execute()
-    if err != nil {
-        fmt.Fprintf(os.Stderr, "Error when calling `NodesApi.GetNodes``: %v\n", err)
-        fmt.Fprintf(os.Stderr, "Full HTTP response: %v\n", r)
+    // Nodes
+    fmt.Println("--- nodes ---")
+    nodes, _, _ := client.NodesApi.GetNodes(ctx).Execute()
+    fmt.Fprintf(os.Stdout, "First hostname: %v\n", nodes.KnownNodes[0].Hostname)
+    capa := int64(1000000000)
+    change := []garage.NodeRoleChange{
+	    garage.NodeRoleChange{NodeRoleUpdate: &garage.NodeRoleUpdate {
+	        Id: *nodes.KnownNodes[0].Id,
+	        Zone: "dc1",
+	        Capacity: *garage.NewNullableInt64(&capa),
+	        Tags: []string{ "fast", "amd64" },
+	    }},
    }
+    staged, _, _ := client.LayoutApi.AddLayout(ctx).NodeRoleChange(change).Execute()
+    msg, _, _ := client.LayoutApi.ApplyLayout(ctx).LayoutVersion(*garage.NewLayoutVersion(staged.Version + 1)).Execute()
+    fmt.Printf(strings.Join(msg.Message, "\n")) // Layout configured

-    // Process the response
-    fmt.Fprintf(os.Stdout, "Target hostname: %v\n", resp.KnownNodes[resp.Node].Hostname)
+    health, _, _ := client.NodesApi.GetHealth(ctx).Execute()
+    fmt.Printf("Status: %s, nodes: %v/%v, storage: %v/%v, partitions: %v/%v\n", health.Status, health.ConnectedNodes, health.KnownNodes, health.StorageNodesOk, health.StorageNodes, health.PartitionsAllOk, health.Partitions)
+
+    // Key
+    fmt.Println("\n--- key ---")
+    key := "openapi-key"
+    keyInfo, _, _ := client.KeyApi.AddKey(ctx).AddKeyRequest(garage.AddKeyRequest{Name: *garage.NewNullableString(&key) }).Execute()
+    defer client.KeyApi.DeleteKey(ctx).Id(*keyInfo.AccessKeyId).Execute()
+    fmt.Printf("AWS_ACCESS_KEY_ID=%s\nAWS_SECRET_ACCESS_KEY=%s\n", *keyInfo.AccessKeyId, *keyInfo.SecretAccessKey.Get())
+
+    id := *keyInfo.AccessKeyId
+    canCreateBucket := true
+    updateKeyRequest := *garage.NewUpdateKeyRequest()
+    updateKeyRequest.SetName("openapi-key-updated")
+    updateKeyRequest.SetAllow(garage.UpdateKeyRequestAllow { CreateBucket: &canCreateBucket })
+    update, _, _ := client.KeyApi.UpdateKey(ctx).Id(id).UpdateKeyRequest(updateKeyRequest).Execute()
+    fmt.Printf("Updated %v with key name %v\n", *update.AccessKeyId, *update.Name)
+
+    keyList, _, _ := client.KeyApi.ListKeys(ctx).Execute()
+    fmt.Printf("Keys count: %v\n", len(keyList))
+
+    // Bucket
+    fmt.Println("\n--- bucket ---")
+    global_name := "global-ns-openapi-bucket"
+    local_name := "local-ns-openapi-bucket"
+    bucketInfo, _, _ := client.BucketApi.CreateBucket(ctx).CreateBucketRequest(garage.CreateBucketRequest{
+        GlobalAlias: &global_name,
+        LocalAlias: &garage.CreateBucketRequestLocalAlias {
+            AccessKeyId: keyInfo.AccessKeyId,
+	    Alias: &local_name,
+        },
+    }).Execute()
+    defer client.BucketApi.DeleteBucket(ctx).Id(*bucketInfo.Id).Execute()
+    fmt.Printf("Bucket id: %s\n", *bucketInfo.Id)
+
+    updateBucketRequest := *garage.NewUpdateBucketRequest()
+    website := garage.NewUpdateBucketRequestWebsiteAccess()
+    website.SetEnabled(true)
+    website.SetIndexDocument("index.html")
+    website.SetErrorDocument("errors/4xx.html")
+    updateBucketRequest.SetWebsiteAccess(*website)
+    quotas := garage.NewUpdateBucketRequestQuotas()
+    quotas.SetMaxSize(1000000000)
+    quotas.SetMaxObjects(999999999)
+    updateBucketRequest.SetQuotas(*quotas)
+    updatedBucket, _, _ := client.BucketApi.UpdateBucket(ctx).Id(*bucketInfo.Id).UpdateBucketRequest(updateBucketRequest).Execute()
+    fmt.Printf("Bucket %v website activation: %v\n", *updatedBucket.Id, *updatedBucket.WebsiteAccess)
+
+    bucketList, _, _ := client.BucketApi.ListBuckets(ctx).Execute()
+    fmt.Printf("Bucket count: %v\n", len(bucketList))
 }
 ```

@@ -31,9 +31,9 @@ npm install --save git+https://git.deuxfleurs.fr/garage-sdk/garage-admin-sdk-js.
 A short example:

 ```javascript
-const garage = require('garage_administration_api_v0garage_v0_8_0');
+const garage = require('garage_administration_api_v1garage_v0_9_0');

-const api = new garage.ApiClient("http://127.0.0.1:3903/v0");
+const api = new garage.ApiClient("http://127.0.0.1:3903/v1");
 api.authentications['bearerAuth'].accessToken = "s3cr3t";

 const [node, layout, key, bucket] = [
@@ -5,13 +5,13 @@ weight = 99

 ## S3

-If you are developping a new application, you may want to use Garage to store your user's media.
+If you are developing a new application, you may want to use Garage to store your user's media.

 The S3 API that Garage uses is a standard REST API, so as long as you can make HTTP requests,
 you can query it. You can check the [S3 REST API Reference](https://docs.aws.amazon.com/AmazonS3/latest/API/API_Operations_Amazon_Simple_Storage_Service.html) from Amazon to learn more.

-Developping your own wrapper around the REST API is time consuming and complicated.
-Instead, there are some libraries already avalaible.
+Developing your own wrapper around the REST API is time consuming and complicated.
+Instead, there are some libraries already available.

 Some of them are maintained by Amazon, some by Minio, others by the community.

@@ -23,7 +23,7 @@ client = minio.Minio(
  "GKyourapikey",
  "abcd[...]1234",
  # Force the region, this is specific to garage
-  region="region",
+  region="garage",
 )
 ```

@@ -80,7 +80,7 @@ from garage_admin_sdk.apis import *
 from garage_admin_sdk.models import *

 configuration = garage_admin_sdk.Configuration(
-  host = "http://localhost:3903/v0",
+  host = "http://localhost:3903/v1",
  access_token = "s3cr3t"
 )

@@ -94,13 +94,14 @@ print(f"running garage {status.garage_version}, node_id {status.node}")

 # Change layout of this node
 current = layout.get_layout()
-layout.add_layout({
-  status.node: NodeClusterInfo(
+layout.add_layout([
+  NodeRoleChange(
+    id = status.node,
    zone = "dc1",
-    capacity = 1,
+    capacity = 1000000000,
    tags = [ "dev" ],
  )
-})
+])
 layout.apply_layout(LayoutVersion(
  version = current.version + 1
 ))
@@ -23,7 +23,7 @@ To configure S3-compatible software to interact with Garage,
 you will need the following parameters:

 - An **API endpoint**: this corresponds to the HTTP or HTTPS address
-  used to contact the Garage server. When runing Garage locally this will usually
+  used to contact the Garage server. When running Garage locally this will usually
  be `http://127.0.0.1:3900`. In a real-world setting, you would usually have a reverse-proxy
  that adds TLS support and makes your Garage server available under a public hostname
  such as `https://garage.example.com`.
@@ -12,8 +12,9 @@ In this section, we cover the following web applications:
 | [Mastodon](#mastodon)     | ✅       | Natively supported    |
 | [Matrix](#matrix)     | ✅       |  Tested with `synapse-s3-storage-provider`    |
 | [ejabberd](#ejabberd)     | ✅       |  `mod_s3_upload`    |
-| [Pixelfed](#pixelfed)     | ❓       |  Not yet tested    |
-| [Pleroma](#pleroma)     | ❓       |  Not yet tested    |
+| [Ente](#ente)     | ✅       |  Natively supported |
+| [Pixelfed](#pixelfed)     | ❓       |  Natively supported    |
+| [Pleroma](#pleroma)     | ✅        |  Natively supported    |
 | [Lemmy](#lemmy)     | ✅        |  Supported with pict-rs    |
 | [Funkwhale](#funkwhale)     | ❓       | Not yet tested     |
 | [Misskey](#misskey)     | ❓       | Not yet tested     |
@@ -37,7 +38,7 @@ Second, we suppose you have created a key and a bucket.
 As a reminder, you can create a key for your nextcloud instance as follow:

 ```bash
-garage key new --name nextcloud-key
+garage key create nextcloud-key
 ```

 Keep the Key ID and the Secret key in a pad, they will be needed later.  
@@ -53,7 +54,7 @@ garage bucket allow nextcloud --read --write --key nextcloud-key

 Now edit your Nextcloud configuration file to enable object storage.
 On my installation, the config. file is located at the following path: `/var/www/nextcloud/config/config.php`.  
-We will add a new root key to the `$CONFIG` dictionnary named `objectstore`:
+We will add a new root key to the `$CONFIG` dictionary named `objectstore`:

 ```php
 <?php
@@ -69,7 +70,7 @@ $CONFIG = array(
        'hostname' => '127.0.0.1', // Can also be a domain name, eg. garage.example.com
        'port' => 3900,            // Put your reverse proxy port or your S3 API port
        'use_ssl' => false,        // Set it to true if you have a TLS enabled reverse proxy
-        'region' => 'garage',      // Garage has only one region named "garage"
+        'region' => 'garage',      // Garage default region is named "garage", edit according to your cluster config
        'use_path_style' => true   // Garage supports only path style, must be set to true
    ],
 ],
@@ -80,6 +81,53 @@ To test your new configuration, just reload your Nextcloud webpage and start sen

 *External link:* [Nextcloud Documentation > Primary Storage](https://docs.nextcloud.com/server/latest/admin_manual/configuration_files/primary_storage.html)

+#### SSE-C encryption (since Garage v1.0)
+
+Since version 1.0, Garage supports server-side encryption with customer keys
+(SSE-C).  In this mode, Garage is responsible for encrypting and decrypting
+objects, but it does not store the encryption key itself. The encryption key
+should be provided by Nextcloud upon each request.  This mode of operation is
+supported by Nextcloud and it has successfully been tested together with
+Garage.
+
+To enable SSE-C encryption:
+
+1. Make sure your Garage server is accessible via SSL through a reverse proxy
+   such as Nginx, and that it is using a valid public certificate (Nextcloud
+   might be able to connect to an S3 server that is using a self-signed
+   certificate, but you will lose many hours while trying, so don't).
+   Configure values for `use_ssl` and `port` accordingly in your `config.php`
+   file.
+
+2. Generate an encryption key using the following command:
+
+    ```
+    openssl rand -base64 32
+    ```
+
+    Make sure to keep this key **secret**!
+
+3. Add the encryption key in your `config.php` file as follows:
+
+
+    ```php
+    <?php
+    $CONFIG = array(
+    'objectstore' => [
+        'class' => '\\OC\\Files\\ObjectStore\\S3',
+        'arguments' => [
+            ...
+            'sse_c_key' => 'exampleencryptionkeyLbU+5fKYQcVoqnn+RaIOXgo=',
+            ...
+        ],
+    ],
+    ```
+
+Nextcloud will now make Garage encrypt files at rest in the storage bucket.
+These files will not be readable by an S3 client that has credentials to the
+bucket but doesn't also know the secret encryption key.
+
+
 ### External Storage

 **From the GUI.** Activate the "External storage support" app from the "Applications" page (click on your account icon on the top right corner of your screen to display the menu). Go to your parameters page (also located below your account icon). Click on external storage (or the corresponding translation in your language).
@@ -88,7 +136,7 @@ To test your new configuration, just reload your Nextcloud webpage and start sen
 *Click on the picture to zoom*

 Add a new external storage. Put what you want in "folder name" (eg. "shared"). Select "Amazon S3". Keep "Access Key" for the Authentication field.
-In Configuration, put your bucket name (eg. nextcloud), the host (eg. 127.0.0.1), the port (eg. 3900 or 443), the region (garage). Tick the SSL box if you have put an HTTPS proxy in front of garage. You must tick the "Path access" box and you must leave the "Legacy authentication (v2)" box empty. Put your Key ID (eg. GK...) and your Secret Key in the last two input boxes. Finally click on the tick symbol on the right of your screen.
+In Configuration, put your bucket name (eg. nextcloud), the host (eg. 127.0.0.1), the port (eg. 3900 or 443), the region ("garage" if you use the default, or the one your configured in your `garage.toml`). Tick the SSL box if you have put an HTTPS proxy in front of garage. You must tick the "Path access" box and you must leave the "Legacy authentication (v2)" box empty. Put your Key ID (eg. GK...) and your Secret Key in the last two input boxes. Finally click on the tick symbol on the right of your screen.

 Now go to your "Files" app and a new "linked folder" has appeared with the name you chose earlier (eg. "shared").

@@ -139,15 +187,15 @@ a reasonable trade-off for some instances.
 Create a key for Peertube:

 ```bash
-garage key new --name peertube-key
+garage key create peertube-key
 ```

 Keep the Key ID and the Secret key in a pad, they will be needed later.  

-We need two buckets, one for normal videos (named peertube-video) and one for webtorrent videos (named peertube-playlist).
+We need two buckets, one for normal videos (named peertube-videos) and one for webtorrent videos (named peertube-playlists).
 ```bash
-garage bucket create peertube-video
-garage bucket create peertube-playlist
+garage bucket create peertube-videos
+garage bucket create peertube-playlists
 ```

 Now we allow our key to read and write on these buckets:
@@ -191,7 +239,7 @@ object_storage:
  # Put localhost only if you have a garage instance running on that node
  endpoint: 'http://localhost:3900' # or "garage.example.com" if you have TLS on port 443

-  # Garage supports only one region for now, named garage
+  # Garage default region is named "garage", edit according to your config
  region: 'garage'

  credentials:
@@ -206,7 +254,7 @@ object_storage:
    proxify_private_files: false

  streaming_playlists:
-    bucket_name: 'peertube-playlist'
+    bucket_name: 'peertube-playlists'

    # Keep it empty for our example
    prefix: ''
@@ -216,7 +264,7 @@ object_storage:

  # Same settings but for webtorrent videos
  videos:
-    bucket_name: 'peertube-video'
+    bucket_name: 'peertube-videos'
    prefix: ''
    # You must fill this field to make Peertube use our reverse proxy/website logic
    base_url: 'http://peertube-videos.web.garage.localhost'
@@ -245,7 +293,7 @@ with average object size ranging from 50 KB to 150 KB.
 As such, your Garage cluster should be configured appropriately for good performance:

 - use Garage v0.8.0 or higher with the [LMDB database engine](@documentation/reference-manual/configuration.md#db-engine-since-v0-8-0).
-  With the default Sled database engine, your database could quickly end up taking tens of GB of disk space.
+  Older versions of Garage used the Sled database engine which had issues, such as databases quickly ending up taking tens of GB of disk space.
 - the Garage database should be stored on a SSD

 ### Creating your bucket
@@ -253,7 +301,7 @@ As such, your Garage cluster should be configured appropriately for good perform
 This is the usual Garage setup:

 ```bash
-garage key new --name mastodon-key
+garage key create mastodon-key
 garage bucket create mastodon-data
 garage bucket allow mastodon-data --read --write --key mastodon-key
 ```
@@ -288,6 +336,7 @@ From the [official Mastodon documentation](https://docs.joinmastodon.org/admin/t

 ```bash
 $ RAILS_ENV=production bin/tootctl media remove --days 3
+$ RAILS_ENV=production bin/tootctl media remove --days 15 --prune-profiles
 $ RAILS_ENV=production bin/tootctl media remove-orphans
 $ RAILS_ENV=production bin/tootctl preview_cards remove --days 15
 ```
@@ -306,8 +355,6 @@ Imports:	1.7 KB
 Settings:	0 Bytes
 ```

-Unfortunately, [old avatars and headers cannot currently be cleaned up](https://github.com/mastodon/mastodon/issues/9567).
-
 ### Migrating your data

 Data migration should be done with an efficient S3 client.
@@ -366,7 +413,7 @@ mc mirror --newer-than "3h" ./public/system/ garage/mastodon-data

 ## Matrix

-Matrix is a chat communication protocol. Its main stable server implementation, [Synapse](https://matrix-org.github.io/synapse/latest/), provides a module to store media on a S3 backend. Additionally, a server independent media store supporting S3 has been developped by the community, it has been made possible thanks to how the matrix API has been designed and will work with implementations like Conduit, Dendrite, etc.
+Matrix is a chat communication protocol. Its main stable server implementation, [Synapse](https://matrix-org.github.io/synapse/latest/), provides a module to store media on a S3 backend. Additionally, a server independent media store supporting S3 has been developed by the community, it has been made possible thanks to how the matrix API has been designed and will work with implementations like Conduit, Dendrite, etc.

 ### synapse-s3-storage-provider (synapse only)

@@ -379,7 +426,7 @@ Supposing you have a working synapse installation, you can add the module with p
 Now create a bucket and a key for your matrix instance (note your Key ID and Secret Key somewhere, they will be needed later):

 ```bash
-garage key new --name matrix-key
+garage key create matrix-key
 garage bucket create matrix
 garage bucket allow matrix --read --write --key matrix-key
 ```
@@ -395,7 +442,7 @@ media_storage_providers:
  store_synchronous: True  # do we want to wait that the file has been written before returning?
  config:
    bucket: matrix       # the name of our bucket, we chose matrix earlier
-    region_name: garage  # only "garage" is supported for the region field
+    region_name: garage  # "garage" by default, edit according to your cluster config
    endpoint_url: http://localhost:3900 # the path to the S3 endpoint
    access_key_id: "GKxxx" # your Key ID
    secret_access_key: "xxxx" # your Secret Key
@@ -403,7 +450,7 @@ media_storage_providers:

 Note that uploaded media will also be stored locally and this behavior can not be deactivated, it is even required for
 some operations like resizing images.
-In fact, your local filesysem is considered as a cache but without any automated way to garbage collect it.
+In fact, your local filesystem is considered as a cache but without any automated way to garbage collect it.

 We can build our garbage collector with `s3_media_upload`, a tool provided with the module.
 If you installed the module with the command provided before, you should be able to bring it in your path:
@@ -521,13 +568,186 @@ The module can then be configured with:
 Other configuration options can be found in the
 [configuration YAML file](https://github.com/processone/ejabberd-contrib/blob/master/mod_s3_upload/conf/mod_s3_upload.yml).

+
+## Ente
+
+Ente is an alternative for Google Photos and Apple Photos. It [can be selfhosted](https://help.ente.io/self-hosting/) and is working fine with Garage as of May 2024.
+As a first step we need to create a bucket and a key for Ente:
+
+```bash
+garage bucket create ente
+garage key create ente-key
+# For the CORS setup to work, the key needs to be --owner as well, at least temporarily.
+garage bucket allow ente --read --write --owner --key ente-key
+```
+
+We also need to setup some CORS rules to allow the Ente frontend to access the bucket:
+
+```bash
+export CORS='{"CORSRules":[{"AllowedHeaders":["*"],"AllowedMethods":["GET", "PUT", "POST", "DELETE"],"AllowedOrigins":["*"], "ExposeHeaders":["ETag"]}]}'
+aws s3api put-bucket-cors --bucket ente --cors-configuration $CORS
+```
+
+Now we need to configure ente-server to use our bucket. This is explained [in the Ente S3 documentation](https://help.ente.io/self-hosting/guides/external-s3).
+Prepare a configuration file for ente's backend as `museum.yaml`:
+
+```yaml
+credentials-file: /credentials.yaml
+apps:
+  public-albums: https://albums.example.tld # If you want to use the share album feature
+    internal:
+hardcoded-ott:
+  local-domain-suffix: "@example.com" # Your domain
+  local-domain-value: 123456 # Custom One-Time Password since we are not sending mail by default
+key:
+  # WARNING -- You MUST CHANGE the values below
+  # Someone has made an image that can do it for you : https://github.com/EdyTheCow/ente-selfhost/blob/main/images/ente-server-tools/Dockerfile
+  # Simply build it yourself or run docker run --rm ghcr.io/edythecow/ente-server-tools go run tools/gen-random-keys/main.go
+  encryption: yvmG/RnzKrbCb9L3mgsmoxXr9H7i2Z4qlbT0mL3ln4w= # CHANGE THIS VALUE
+  hash: KXYiG07wC7GIgvCSdg+WmyWdXDAn6XKYJtp/wkEU7x573+byBRAYtpTP0wwvi8i/4l37uicX1dVTUzwH3sLZyw== # CHANGE THIS VALUE
+jwt:
+  secret: i2DecQmfGreG6q1vBj5tCokhlN41gcfS2cjOs9Po-u8= # CHANGE THIS VALUE
+```
+
+The full configuration file can be found [here](https://github.com/ente-io/ente/blob/main/server/configurations/local.yaml)
+Then prepare a credentials file as `credentials.yaml`
+
+```yaml
+db:
+    host: postgres
+    port: 5432
+    name: <ente_db_name>
+    user: <pguser>
+    password: <pgpass>
+
+s3:
+    # Override the primary and secondary hot storage. The commented out values
+    # are the defaults.
+    #
+    hot_storage:
+        primary: b2-eu-cen
+    #    secondary: wasabi-eu-central-2-v3
+
+    # If true, enable some workarounds to allow us to use a local minio instance
+    # for object storage.
+    #
+    # 1. Disable SSL.
+    # 2. Use "path" style S3 URLs (see `use_path_style_urls` below).
+    # 3. Directly download the file during replication instead of going via the
+    #    Cloudflare worker.
+    # 4. Do not specify storage classes when uploading objects (since minio does
+    #    not support them, specifically it doesn't support GLACIER).
+    are_local_buckets: true
+
+    # To use "path" style S3 URLs instead of DNS-based bucket access
+    # default to true if you set "are_local_buckets: true"
+    # use_path_style_urls: true
+
+    b2-eu-cen: # Don't change this key, it is hardcoded
+        key: <keyID> 
+        secret: <keySecret>
+        endpoint: garage:3900 # publicly accessible endpoint of your garage instance
+        region: garage
+        bucket: <yourbucketName>
+        use_path_style: true
+    # you can specify secondary locations, names are hardcoded as well
+    # wasabi-eu-central-2-v3:
+    # scw-eu-fr-v3:
+    
+    # and you can also specify a bucket to be used for embeddings, preview etc..
+    # default to the first bucket 
+    # derived-storage: wasabi-eu-central-2-derived
+```
+
+Finally you can run it with Docker : 
+
+```bash
+docker run -d --name ente-server --restart unless-stopped -v /path/to/museum.yaml:/museum.yaml -v /path/to/credentials.yaml:/credentials.yaml -p 8080:8080 ghcr.io/ente-io/ente-server
+```
+
+For more information on deployment you can check the [ente documentation](https://help.ente.io/self-hosting/)
+
 ## Pixelfed

 [Pixelfed Technical Documentation > Configuration](https://docs.pixelfed.org/technical-documentation/env.html#filesystem)

 ## Pleroma

-[Pleroma Documentation > Pleroma.Uploaders.S3](https://docs-develop.pleroma.social/backend/configuration/cheatsheet/#pleromauploaderss3)
+### Creating your bucket
+
+This is the usual Garage setup:
+
+```bash
+garage key new --name pleroma-key
+garage bucket create pleroma
+garage bucket allow pleroma --read --write --owner --key pleroma-key
+```
+
+We also need to expose these buckets publicly to serve their content to users:
+
+```bash
+garage bucket website --allow pleroma
+```
+
+Note the Key ID and Secret Key.
+
+### Configure Pleroma
+
+Update your Pleroma configuration like that in `/etc/pleroma/config.exs`.
+
+```
+config :pleroma, Pleroma.Upload,
+    uploader: Pleroma.Uploaders.S3,
+    base_url: "https://pleroma.garage.example.tld"
+
+config :ex_aws, :s3,
+    access_key_id: "GW...",
+    secret_access_key: "XXX",
+    region: "garage",
+    host: "api.garage.example.tld"
+```
+
+And restart Pleroma.
+
+You can found more information in [Pleroma Documentation > Pleroma.Uploaders.S3](https://docs-develop.pleroma.social/backend/configuration/cheatsheet/#pleromauploaderss3)
+
+### Migrating your data
+
+Pleroma have an internal migration tool that can encounter some fatal error
+
+```
+** (EXIT from #PID<0.98.0>) an exception was raised:
+    ** (File.Error) could not stream "/var/lib/pleroma/uploads/09/f8": illegal operation on a directory
+        (elixir 1.17.3) lib/file/stream.ex:100: anonymous fn/3 in Enumerable.File.Stream.reduce/3
+        (elixir 1.17.3) lib/stream.ex:1675: anonymous fn/5 in Stream.resource/3
+        (elixir 1.17.3) lib/stream.ex:1891: Enumerable.Stream.do_each/4
+        (elixir 1.17.3) lib/task/supervised.ex:370: Task.Supervised.stream_reduce/7
+        (elixir 1.17.3) lib/enum.ex:4423: Enum.map/2
+        (ex_aws_s3 2.5.8) lib/ex_aws/s3/upload.ex:141: ExAws.Operation.ExAws.S3.Upload.perform/2
+        (pleroma 2.10.0) lib/pleroma/uploaders/s3.ex:60: Pleroma.Uploaders.S3.put_file/1
+        (pleroma 2.10.0) lib/pleroma/uploaders/uploader.ex:49: Pleroma.Uploaders.Uploader.put_file/2
+```
+
+So, use [your best tool](https://garagehq.deuxfleurs.fr/documentation/connect/cli/) to sync `/var/lib/pleroma/uploads/` in your S3.
+
+Then, to avoid some non existent problem (just in case of), run this command
+
+```bash
+while true
+do
+    rm -vr $(./bin/pleroma_ctl uploads migrate_local S3 2>&1 | grep "could not stream" | awk -F '"' '{print $2}')
+    sleep 5
+done
+```
+
+If you have many files, stop this command sometime and the command bellow (interactive) to delete local
+file after upload. Then restart the loop.
+
+```bash
+./bin/pleroma_ctl uploads migrate_local S3 --delete
+```
+
+And *voilà*

 ## Lemmy

@@ -54,9 +54,9 @@ how to configure this.
 Create your key and bucket:

 ```bash
-garage key new my-key
-garage bucket create backup
-garage bucket allow backup --read --write --key my-key
+garage key create my-key
+garage bucket create backups
+garage bucket allow backups --read --write --key my-key
 ```

 Then register your Key ID and Secret key in your environment:
@@ -161,3 +161,59 @@ kopia repository validate-provider

 You can then run all the standard kopia commands: `kopia snapshot create`, `kopia mount`...
 Everything should work out-of-the-box.
+
+## Plakar
+
+Create your key and bucket on Garage server:
+
+```bash
+garage key create my-plakar-key
+garage bucket create plakar-backups
+garage bucket allow plakar-backups --read --write --key my-plakar-key
+…
+```
+
+On Plakar server, add your Garage as a storage location:
+```bash
+plakar store add garageS3 s3://my-garage.tld/plakar-backups \
+region=garage # Or as you've specified in garage.toml \
+access_key=<Key ID from "garage key info my-plakar-key"> \
+secret_access_key=<Secret key from "garage key info my-plakar-key">
+```
+
+Then create the repository.
+```bash
+plakar at @garageS3 create -plaintext # Unencrypted
+# or
+plakar at @garageS3 create #encrypted
+```
+
+If you encrypt your backups (Plakar default), you will need to define a strong passphrase. Do not forget to save your password safely. It will be needed to decrypt your backups.
+
+
+After the repository has been created, check that everything works as expected (that might give an empty result as no file has been added yet, but no error message):
+```bash
+plakar at @garageS3 check
+```
+
+Now that everything is configure, you can use Garage as your backups storage. For instance sync it with a local backup storage:
+```bash
+$ plakar at ~/backups sync to @garageS3
+```
+
+Or list the S3 storage content:
+```bash
+$ plakar at @garageS3 ls
+```
+
+More information in Plakar documentation: https://www.plakar.io/docs/main/quickstart/
+
+## Synology HyperBackup
+
+HyperBackup can be configured to upload backups to garage using a custom S3 destination. However, the HyperBackup client hardcodes the `us-east-1` region that is a critical input to the v4 signature process. If garage is not set to `us-east-1`, HyperBackup will recognize available buckets, but fail during the final setup stage.
+
+In garage.toml:
+```toml
+[s3_api]
+s3_region = "us-east-1"
+```
@@ -41,7 +41,7 @@ Some commands:
 # list buckets
 mc ls garage/

-# list objets in a bucket
+# list objects in a bucket
 mc ls garage/my_files

 # copy from your filesystem to garage
@@ -149,6 +149,15 @@ rclone help
 This will tremendously accelerate operations such as `rclone sync` or `rclone ncdu` by reducing the number
 of ListObjects calls that are made.

+**Garage behind Cloudflare proxy:** when running Garage behind Cloudflare proxy, you might see `Response: error 403 Forbidden, Forbidden: Invalid signature` error in your garage logs or `AccessDenied: Forbidden: Invalid signature` error in rclone logs. Try adding `--s3-sign-accept-encoding=false` flag to your rclone command and see if the issue is resolved.
+
+```bash
+# this throws an error
+rclone lsd garage:
+
+# this should work
+rclone lsd --s3-sign-accept-encoding=false garage:
+```

 ## `s3cmd`

@@ -209,7 +218,7 @@ Within Cyberduck, a
 available within the `Preferences -> Profiles` section. This can enabled and
 then connections to Garage may be configured.

-### Instuctions for the CLI
+### Instructions for the CLI

 To configure duck (Cyberduck's CLI tool), start by creating its folder hierarchy:

@@ -259,7 +268,7 @@ duck --delete garage:/my-files/an-object.txt

 ## WinSCP (libs3) {#winscp}

-*You can find instructions on how to use the GUI in french [in our wiki](https://wiki.deuxfleurs.fr/fr/Guide/Garage/WinSCP).*
+*You can find instructions on how to use the GUI in french [in our wiki](https://guide.deuxfleurs.fr/prise_en_main/winscp/).*

 How to use `winscp.com`, the CLI interface of WinSCP:

@@ -314,4 +323,3 @@ ls
 ```

 And through the web interface at http://[::1]:8080/web/client
-
@@ -17,13 +17,13 @@ Garage can also help you serve this content.

 ## Gitea

-You can use Garage with Gitea to store your [git LFS](https://git-lfs.github.com/) data, your users' avatar, and their attachements.
+You can use Garage with Gitea to store your [git LFS](https://git-lfs.github.com/) data, your users' avatar, and their attachments.
 You can configure a different target for each data type (check `[lfs]` and `[attachment]` sections of the Gitea documentation) and you can provide a default one through the `[storage]` section.

 Let's start by creating a key and a bucket (your key id and secret will be needed later, keep them somewhere):

 ```bash
-garage key new --name gitea-key
+garage key create gitea-key
 garage bucket create gitea
 garage bucket allow gitea --read --write --key gitea-key
 ```
@@ -118,7 +118,7 @@ through another support, like a git repository.
 As a first step, we will need to create a bucket on Garage and enabling website access on it:

 ```bash
-garage key new --name nix-key
+garage key create nix-key
 garage bucket create nix.example.com
 garage bucket allow nix.example.com --read --write --key nix-key
 garage bucket website nix.example.com --allow
@@ -201,11 +201,9 @@ on the binary cache, the client will download the result from the cache instead

 ### Channels

-Channels additionnaly serve Nix definitions, ie. a `.nix` file referencing
+Channels additionally serve Nix definitions, ie. a `.nix` file referencing
 all the derivations you want to serve.

 ## Gitlab

 *External link:* [Gitlab Documentation > Object storage](https://docs.gitlab.com/ee/administration/object_storage.html)
-
-
@@ -8,18 +8,18 @@ have published Ansible roles.  We list them and compare them below.

 ## Comparison of Ansible roles

-| Feature                            | [ansible-role-garage](#zorun-ansible-role-garage) | [garage-docker-ansible-deploy](#moan0s-garage-docker-ansible-deploy) |
-|------------------------------------|---------------------------------------------|---------------------------------------------------------------|
-| **Runtime**                        | Systemd                                     | Docker                                                        |
-| **Target OS**                      | Any Linux                                   | Any Linux                                                     |
-| **Architecture**                   | amd64, arm64, i686                          | amd64, arm64                                                  |
-| **Additional software**            | None                                        | Traefik                                                       |
-| **Automatic node connection**      | ❌                                          | ✅                                                            |
-| **Layout management**              | ❌                                          | ✅                                                            |
-| **Manage buckets & keys**          | ❌                                          | ✅ (basic)                                                    |
-| **Allow custom Garage config**     | ✅                                          | ❌                                                            |
-| **Facilitate Garage upgrades**     | ✅                                          | ❌                                                            |
-| **Multiple instances on one host** | ✅                                          | ✅                                                            |
+| Feature                            | [ansible-role-garage](#zorun-ansible-role-garage) | [garage-docker-ansible-deploy](#moan0s-garage-docker-ansible-deploy) | [eddster2309 ansible-role-garage](#eddster2309-ansible-role-garage) |
+|------------------------------------|---------------------------------------------|---------------------------------------------------------------|---------------------------------|
+| **Runtime**                        | Systemd                                     | Docker                                                        | Systemd                         |
+| **Target OS**                      | Any Linux                                   | Any Linux                                                     | Any Linux                       |
+| **Architecture**                   | amd64, arm64, i686                          | amd64, arm64                                                  | arm64, arm, 386, amd64                            |
+| **Additional software**            | None                                        | Traefik                                                       | Nginx and Keepalived (optional) |
+| **Automatic node connection**      | ❌                                          | ✅                                                            | ✅                              |
+| **Layout management**              | ❌                                          | ✅                                                            | ✅                              |
+| **Manage buckets & keys**          | ❌                                          | ✅ (basic)                                                    | ✅                              |
+| **Allow custom Garage config**     | ✅                                          | ❌                                                            | ❌                              |
+| **Facilitate Garage upgrades**     | ✅                                          | ❌                                                            | ✅                              |
+| **Multiple instances on one host** | ✅                                          | ✅                                                            | ❌                              |


 ## zorun/ansible-role-garage
@@ -49,3 +49,15 @@ structured DNS names, etc).

 As a result, this role makes it easier to start with Garage on Ansible,
 but is less flexible.
+
+## eddster2309/ansible-role-garage
+
+[Source code](https://github.com/eddster2309/ansible-role-garage), [Ansible galaxy](https://galaxy.ansible.com/ui/standalone/roles/eddster2309/garage/)
+
+This role is a opinionated but customisable role using the official Garage
+static binaries and only requires Systemd. As such it should work on any
+Linux based host. It includes all the nesscary configuration to
+automatically setup a clustered Garage deployment. Most Garage
+configuration options are exposed through Ansible variables so while you
+can't provide a custom config you can get very close. It can optionally
+installed a HA nginx deployment with Keepalived.
@@ -15,9 +15,10 @@ Alpine Linux repositories (available since v3.17):
 apk add garage
 ```

-The default configuration file is installed to `/etc/garage.toml`. You can run
-Garage using: `rc-service garage start`. If you don't specify `rpc_secret`, it
-will be automatically replaced with a random string on the first start.
+The default configuration file is installed to `/etc/garage/garage.toml`. You can run
+Garage using: `rc-service garage start`.
+
+If you don't specify `rpc_secret`, it will be automatically replaced with a random string on the first start.

 Please note that this package is built without Consul discovery, Kubernetes
 discovery, OpenTelemetry exporter, and K2V features (K2V will be enabled once
@@ -26,7 +27,11 @@ it's stable).

 ## Arch Linux

-Garage is available in the [AUR](https://aur.archlinux.org/packages/garage).
+Garage is available in the official repositories under [extra](https://archlinux.org/packages/extra/x86_64/garage).
+
+```bash
+pacman -S garage
+```

 ## FreeBSD

@@ -39,3 +44,9 @@ pkg install garage
 ```bash
 nix-shell -p garage
 ```
+
+## conda-forge
+
+```bash
+pixi global install garage
+```
@@ -33,7 +33,7 @@ by adding encryption at different levels.

 We would be very curious to know your needs and thougs about ideas such as
 encryption practices and things like key management, as we want Garage to be a
-serious base platform for the developpment of secure, encrypted applications.
+serious base platform for the development of secure, encrypted applications.
 Do not hesitate to come talk to us if you have any thoughts or questions on the
 subject.

@@ -53,20 +53,43 @@ and that's also why your nodes have super long identifiers.

 Adding TLS support built into Garage is not currently planned.

-## Garage stores data in plain text on the filesystem
+## Garage stores data in plain text on the filesystem or encrypted using customer keys (SSE-C)

-Garage does not handle data encryption at rest by itself, and instead delegates
-to the user to add encryption, either at the storage layer (LUKS, etc) or on
-the client side (or both). There are no current plans to add data encryption
-directly in Garage.
+For standard S3 API requests, Garage does not encrypt data at rest by itself.
+For the most generic at rest encryption of data, we recommend setting up your
+storage partitions on encrypted LUKS devices.

-Implementing data encryption directly in Garage might make things simpler for
-end users, but also raises many more questions, especially around key
-management: for encryption of data, where could Garage get the encryption keys
-from ? If we encrypt data but keep the keys in a plaintext file next to them,
-it's useless. We probably don't want to have to manage secrets in garage as it
-would be very hard to do in a secure way. Maybe integrate with an external
-system such as Hashicorp Vault?
+If you are developing your own client software that makes use of S3 storage,
+we recommend implementing data encryption directly on the client side and never
+transmitting plaintext data to Garage. This makes it easy to use an external
+untrusted storage provider if necessary.
+
+Garage does support [SSE-C
+encryption](https://docs.aws.amazon.com/AmazonS3/latest/userguide/ServerSideEncryptionCustomerKeys.html),
+an encryption mode of Amazon S3 where data is encrypted at rest using
+encryption keys given by the client.  The encryption keys are passed to the
+server in a header in each request, to encrypt or decrypt data at the moment of
+reading or writing. The server discards the key as soon as it has finished
+using it for the request.  This mode allows the data to be encrypted at rest by
+Garage itself, but it requires support in the client software. It is also not
+adapted to a model where the server is not trusted or assumed to be
+compromised, as the server can easily know the encryption keys.  Note however
+that when using SSE-C encryption, the only Garage node that knows the
+encryption key passed in a given request is the node to which the request is
+directed (which can be a gateway node), so it is easy to have untrusted nodes
+in the cluster as long as S3 API requests containing SSE-C encryption keys are
+not directed to them.
+
+Implementing automatic data encryption directly in Garage without client-side
+management of keys (something like
+[SSE-S3](https://docs.aws.amazon.com/AmazonS3/latest/userguide/UsingServerSideEncryption.html))
+could make things simpler for end users that don't want to setup LUKS, but also
+raises many more questions, especially around key management: for encryption of
+data, where could Garage get the encryption keys from? If we encrypt data but
+keep the keys in a plaintext file next to them, it's useless. We probably don't
+want to have to manage secrets in Garage as it would be very hard to do in a
+secure way. At the time of speaking, there are no plans to implement this in
+Garage.


 # Adding data encryption using external tools
@@ -85,14 +108,14 @@ Protects against the following threats:

 - Stolen HDD

-Crucially, does not protect againt malicious sysadmins or remote attackers that
+Crucially, does not protect against malicious sysadmins or remote attackers that
 might gain access to your servers.

 Methods include full-disk encryption with tools such as LUKS.

 ## Encrypting data on the client side

-Protects againt the following threats:
+Protects against the following threats:

 - A honest-but-curious administrator
 - A malicious administrator that tries to corrupt your data
@@ -9,7 +9,7 @@ There are three methods to expose buckets as website:

 1. using the PutBucketWebsite S3 API call, which is allowed for access keys that have the owner permission bit set

-2. from the Garage CLI, by an adminstrator of the cluster
+2. from the Garage CLI, by an administrator of the cluster

 3. using the Garage administration API

@@ -38,7 +38,7 @@ Our website serving logic is as follow:

 Now we need to infer the URL of your website through your bucket name.
 Let assume:
-  - we set `root_domain = ".web.example.com"` in `garage.toml` ([ref](@/documentation/reference-manual/configuration.md#root_domain))
+  - we set `root_domain = ".web.example.com"` in `garage.toml` ([ref](@/documentation/reference-manual/configuration.md#web_root_domain))
  - our bucket name is `garagehq.deuxfleurs.fr`.

 Our bucket will be served if the Host field matches one of these 2 values (the port is ignored):
@@ -20,12 +20,12 @@ sudo apt-get update
 sudo apt-get install build-essential
 ```

-## Building from source from the Gitea repository
+## Building from source from the Forgejo repository

 The primary location for Garage's source code is the
-[Gitea repository](https://git.deuxfleurs.fr/Deuxfleurs/garage),
+[Forgejo repository](https://git.deuxfleurs.fr/Deuxfleurs/garage),
 which contains all of the released versions as well as the code
-for the developpement of the next version.
+for the development of the next version.

 Clone the repository and enter it as follows:

@@ -41,7 +41,7 @@ git tag  				# List available tags
 git checkout v0.8.0		# Change v0.8.0 with the version you wish to build
 ```

-Otherwise you will be building a developpement build from the `main` branch
+Otherwise you will be building a development build from the `main` branch
 that includes all of the changes to be released in the next version.
 Be careful that such a build might be unstable or contain bugs,
 and could be incompatible with nodes that run stable versions of Garage.
@@ -85,11 +85,14 @@ The following feature flags are available in v0.8.0:
 | Feature flag | Enabled | Description |
 | ------------ | ------- | ----------- |
 | `bundled-libs` | *by default* | Use bundled version of sqlite3, zstd, lmdb and libsodium |
-| `system-libs` | optional | Use system version of sqlite3, zstd, lmdb and libsodium<br>if available (exclusive with `bundled-libs`, build using<br>`cargo build --no-default-features --features system-libs`) |
+| `consul-discovery` | optional | Enable automatic registration and discovery<br>of cluster nodes through the Consul API |
+| `fjall` | experimental | Enable using Fjall to store Garage's metadata |
+| `journald` | optional | Enable logging to systemd-journald with<br>`GARAGE_LOG_TO_JOURNALD=true` environment variable set |
 | `k2v` | optional | Enable the experimental K2V API (if used, all nodes on your<br>Garage cluster must have it enabled as well) |
 | `kubernetes-discovery` | optional | Enable automatic registration and discovery<br>of cluster nodes through the Kubernetes API |
+| `lmdb` | *by default* | Enable using LMDB to store Garage's metadata |
 | `metrics` | *by default* | Enable collection of metrics in Prometheus format on the admin API |
+| `sqlite` | *by default* | Enable using Sqlite3 to store Garage's metadata |
+| `syslog` | optional | Enable logging to Syslog with<br>`GARAGE_LOG_TO_SYSLOG=true` environment variable set |
+| `system-libs` | optional | Use system version of sqlite3, zstd, lmdb and libsodium<br>if available (exclusive with `bundled-libs`, build using<br>`cargo build --no-default-features --features system-libs`) |
 | `telemetry-otlp` | optional | Enable collection of execution traces using OpenTelemetry |
-| `sled` | *by default* | Enable using Sled to store Garage's metadata |
-| `lmdb` | optional | Enable using LMDB to store Garage's metadata |
-| `sqlite` | optional | Enable using Sqlite3 to store Garage's metadata |
@@ -11,7 +11,7 @@ Firstly clone the repository:

 ```bash
 git clone https://git.deuxfleurs.fr/Deuxfleurs/garage
-cd garage/scripts/helm
+cd garage/script/helm
 ```

 Deploy with default options:
@@ -26,6 +26,13 @@ Or deploy with custom values:
 helm install --create-namespace --namespace garage garage ./garage -f values.override.yaml
 ```

+If you want to manage the CustomResourceDefinition used by garage for its `kubernetes_discovery` outside of the helm chart, add `garage.kubernetesSkipCrd: true` to your custom values and use the kustomization before deploying the helm chart:
+
+```bash
+kubectl apply -k ../k8s/crd
+helm install --create-namespace --namespace garage garage ./garage -f values.override.yaml
+```
+
 After deploying, cluster layout must be configured manually as described in [Creating a cluster layout](@/documentation/quick-start/_index.md#creating-a-cluster-layout). Use the following command to access garage CLI:

 ```bash
@@ -40,12 +47,12 @@ All possible configuration values can be found with:
 helm show values ./garage
 ```

-This is an example `values.overrride.yaml` for deploying in a microk8s cluster with a https s3 api ingress route:
+This is an example `values.override.yaml` for deploying in a microk8s cluster with a https s3 api ingress route:

 ```yaml
 garage:
  # Use only 2 replicas per object
-  replicationMode: "2"
+  replicationFactor: 2

 # Start 4 instances (StatefulSets) of garage
 deployment:
@@ -86,3 +93,62 @@ helm delete --namespace garage garage
 ```

 Note that this will leave behind custom CRD `garagenodes.deuxfleurs.fr`, which must be removed manually if desired.
+
+## Increase PVC size on running Garage instances
+
+Since the Garage Helm chart creates the data and meta PVC based on `StatefulSet` templates, increasing the PVC size can be a bit tricky.
+
+### Confirm the `StorageClass` used for Garage supports volume expansion
+
+Confirm the storage class used for garage.
+
+```bash
+kubectl -n garage get pvc 
+NAME            STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS     VOLUMEATTRIBUTESCLASS   AGE
+data-garage-0   Bound    pvc-080360c9-8ce3-4acf-8579-1701e57b7f3f   30Gi       RWO            longhorn-local   <unset>                 77d
+data-garage-1   Bound    pvc-ab8ba697-6030-4fc7-ab3c-0d6df9e3dbc0   30Gi       RWO            longhorn-local   <unset>                 5d8h
+data-garage-2   Bound    pvc-3ab37551-0231-4604-986d-136d0fd950ec   30Gi       RWO            longhorn-local   <unset>                 5d5h
+meta-garage-0   Bound    pvc-3b457302-3023-4169-846e-c928c5f2ea65   3Gi        RWO            longhorn-local   <unset>                 77d
+meta-garage-1   Bound    pvc-49ace2b9-5c85-42df-9247-51c4cf64b460   3Gi        RWO            longhorn-local   <unset>                 5d8h
+meta-garage-2   Bound    pvc-99e2e50f-42b4-4128-ae2f-b52629259723   3Gi        RWO            longhorn-local   <unset>                 5d5h
+```
+
+In this case, the storage class is `longhorn-local`. Now, check if `ALLOWVOLUMEEXPANSION` is true for the used `StorageClass`.
+
+```bash
+kubectl get storageclasses.storage.k8s.io longhorn-local
+NAME             PROVISIONER          RECLAIMPOLICY   VOLUMEBINDINGMODE   ALLOWVOLUMEEXPANSION   AGE
+longhorn-local   driver.longhorn.io   Delete          Immediate           true                   103d
+```
+
+If your `StorageClass` does not support volume expansion, double check if you can enable it. Otherwise, your only real option is to spin up a new Garage cluster with increased size and migrate all data over.
+
+If your `StorageClass` supports expansion, you are free to continue.
+
+### Increase the size of the PVCs
+
+Increase the size of all PVCs to your desired size.
+
+```bash
+kubectl -n garage edit pvc data-garage-0
+kubectl -n garage edit pvc data-garage-1
+kubectl -n garage edit pvc data-garage-2
+kubectl -n garage edit pvc meta-garage-0
+kubectl -n garage edit pvc meta-garage-1
+kubectl -n garage edit pvc meta-garage-2
+```
+
+### Increase the size of the `StatefulSet` PVC template
+
+This is an optional step, but if not done, future instances of Garage will be created with the original size from the template.
+
+```bash
+kubectl -n garage delete sts --cascade=orphan garage 
+statefulset.apps "garage" deleted
+```
+
+This will remove the Garage `StatefulSet` but leave the pods running. It may seem destructive but needs to be done this way since edits to the size of PVC templates are prohibited.
+
+### Redeploy the `StatefulSet`
+
+Now the size of future PVCs can be increased, and the Garage Helm chart can be upgraded. The new `StatefulSet` should take ownership of the orphaned pods again.
@@ -18,7 +18,7 @@ api_bind_addr = "0.0.0.0:3903"
 ```

 This will allow anyone to scrape Prometheus metrics by fetching
-`http://localhost:3093/metrics`. If you want to restrict access
+`http://localhost:3903/metrics`. If you want to restrict access
 to the exported metrics, set the `metrics_token` configuration value
 to a bearer token to be used when fetching the metrics endpoint.

@@ -19,14 +19,15 @@ To run a real-world deployment, make sure the following conditions are met:

 - You have at least three machines with sufficient storage space available.

- Each machine has a public IP address which is reachable by other machines. It
-  is highly recommended that you use IPv6 for this end-to-end connectivity. If
-  IPv6 is not available, then using a mesh VPN such as
+- Each machine has an IP address which makes it directly reachable by all other machines.
+  In many cases, nodes will be behind a NAT and will not each have a public
+  IPv4 addresses. In this case, is recommended that you use IPv6 for this
+  end-to-end connectivity if it is available. Otherwise, using a mesh VPN such as
  [Nebula](https://github.com/slackhq/nebula) or
  [Yggdrasil](https://yggdrasil-network.github.io/) are approaches to consider
  in addition to building out your own VPN tunneling.

- This guide will assume you are using Docker containers to deploy Garage on each node. 
+- This guide will assume you are using Docker containers to deploy Garage on each node.
  Garage can also be run independently, for instance as a [Systemd service](@/documentation/cookbook/systemd.md).
  You can also use an orchestrator such as Nomad or Kubernetes to automatically manage
  Docker containers on a fleet of nodes.
@@ -42,7 +43,7 @@ For our example, we will suppose the following infrastructure with IPv6 connecti
 | Brussels | Mars    | fc00:F::1  | 1.5 TB     |

 Note that Garage will **always** store the three copies of your data on nodes at different
-locations. This means that in the case of this small example, the available capacity
+locations. This means that in the case of this small example, the usable capacity
 of the cluster is in fact only 1.5 TB, because nodes in Brussels can't store more than that.
 This also means that nodes in Paris and London will be under-utilized.
 To make better use of the available hardware, you should ensure that the capacity
@@ -52,9 +53,9 @@ to store 2 TB of data in total.

 ### Best practices

- If you have fast dedicated networking between all your nodes, and are planing to store
-  very large files, bump the `block_size` configuration parameter to 10 MB
-  (`block_size = 10485760`).
+- If you have reasonably fast networking between all your nodes, and are planing to store
+  mostly large files, bump the `block_size` configuration parameter to 10 MB
+  (`block_size = "10M"`).

 - Garage stores its files in two locations: it uses a metadata directory to store frequently-accessed
  small metadata items, and a data directory to store data blocks of uploaded objects.
@@ -67,36 +68,42 @@ to store 2 TB of data in total.
  EXT4 is not recommended as it has more strict limitations on the number of inodes,
  which might cause issues with Garage when large numbers of objects are stored.

- If you only have an HDD and no SSD, it's fine to put your metadata alongside the data
-  on the same drive. Having lots of RAM for your kernel to cache the metadata will
-  help a lot with performance.  Make sure to use the LMDB database engine,
-  instead of Sled, which suffers from quite bad performance degradation on HDDs.
-  Sled is still the default for legacy reasons, but is not recommended anymore.
+- Servers with multiple HDDs are supported natively by Garage without resorting
+  to RAID, see [our dedicated documentation page](@/documentation/operations/multi-hdd.md).

 - For the metadata storage, Garage does not do checksumming and integrity
-  verification on its own. If you are afraid of bitrot/data corruption,
-  put your metadata directory on a BTRFS partition. Otherwise, just use regular
-  EXT4 or XFS.
+  verification on its own, so it is better to use a robust filesystem such as
+  BTRFS or ZFS. Users have reported that when using the LMDB database engine
+  (the default), database files have a tendency of becoming corrupted after an
+  unclean shutdown (e.g. a power outage), so you should take regular snapshots
+  to be able to recover from such a situation.  This can be done using Garage's
+  built-in automatic snapshotting (since v0.9.4), or by using filesystem level
+  snapshots. If you cannot do so, you might want to switch to Sqlite which is
+  more robust.

- Having a single server with several storage drives is currently not very well
-  supported in Garage ([#218](https://git.deuxfleurs.fr/Deuxfleurs/garage/issues/218)).
-  For an easy setup, just put all your drives in a RAID0 or a ZFS RAIDZ array.
-  If you're adventurous, you can try to format each of your disk as
-  a separate XFS partition, and then run one `garage` daemon per disk drive,
-  or use something like [`mergerfs`](https://github.com/trapexit/mergerfs) to merge
-  all your disks in a single union filesystem that spreads load over them.
+- LMDB is the fastest and most tested database engine, but it has the following
+  weaknesses: 1/ data files are not architecture-independent, you cannot simply
+  move a Garage metadata directory between nodes running different architectures,
+  and 2/ LMDB is not suited for 32-bit platforms. Sqlite is a viable alternative
+  if any of these are of concern.
+
+- If you only have an HDD and no SSD, it's fine to put your metadata alongside
+  the data on the same drive, but then consider your filesystem choice wisely
+  (see above). Having lots of RAM for your kernel to cache the metadata will
+  help a lot with performance.  The default LMDB database engine is the most
+  tested and has good performance.

 ## Get a Docker image

 Our docker image is currently named `dxflrs/garage` and is stored on the [Docker Hub](https://hub.docker.com/r/dxflrs/garage/tags?page=1&ordering=last_updated).
-We encourage you to use a fixed tag (eg. `v0.8.0`) and not the `latest` tag.
-For this example, we will use the latest published version at the time of the writing which is `v0.8.0` but it's up to you
+We encourage you to use a fixed tag (eg. `v2.3.0`) and not the `latest` tag.
+For this example, we will use the latest published version at the time of the writing which is `v2.3.0` but it's up to you
 to check [the most recent versions on the Docker Hub](https://hub.docker.com/r/dxflrs/garage/tags?page=1&ordering=last_updated).

 For example:

 ```
-sudo docker pull dxflrs/garage:v0.8.0
+docker pull dxflrs/garage:v2.3.0
 ```

 ## Deploying and configuring Garage
@@ -119,8 +126,9 @@ A valid `/etc/garage.toml` for our cluster would look as follows:
 metadata_dir = "/var/lib/garage/meta"
 data_dir = "/var/lib/garage/data"
 db_engine = "lmdb"
+metadata_auto_snapshot_interval = "6h"

-replication_mode = "3"
+replication_factor = 3

 compression_level = 2

@@ -144,6 +152,8 @@ Check the following for your configuration files:
 - Make sure `rpc_public_addr` contains the public IP address of the node you are configuring.
  This parameter is optional but recommended: if your nodes have trouble communicating with
  one another, consider adding it.
+  Alternatively, you can also set `rpc_public_addr_subnet`, which can filter
+  the addresses announced to other peers to a specific subnet.

 - Make sure `rpc_secret` is the same value on all nodes. It should be a 32-bytes hex-encoded secret key.
  You can generate such a key with `openssl rand -hex 32`.
@@ -161,12 +171,13 @@ docker run \
  -v /etc/garage.toml:/etc/garage.toml \
  -v /var/lib/garage/meta:/var/lib/garage/meta \
  -v /var/lib/garage/data:/var/lib/garage/data \
-  dxflrs/garage:v0.8.0
+  dxflrs/garage:v2.3.0
 ```

-It should be restarted automatically at each reboot.
-Please note that we use host networking as otherwise Docker containers
-can not communicate with IPv6.
+With this command line, Garage should be started automatically at each boot.
+Please note that we use host networking as otherwise the network indirection
+added by Docker would prevent Garage nodes from communicating with one another
+(especially if using IPv6).

 If you want to use `docker-compose`, you may use the following `docker-compose.yml` file as a reference:

@@ -174,7 +185,7 @@ If you want to use `docker-compose`, you may use the following `docker-compose.y
 version: "3"
 services:
  garage:
-    image: dxflrs/garage:v0.8.0
+    image: dxflrs/garage:v2.3.0
    network_mode: "host"
    restart: unless-stopped
    volumes:
@@ -183,12 +194,14 @@ services:
      - /var/lib/garage/data:/var/lib/garage/data
 ```

-Upgrading between Garage versions should be supported transparently,
-but please check the relase notes before doing so!
-To upgrade, simply stop and remove this container and
-start again the command with a new version of Garage.
+If you wish to upgrade your cluster, make sure to read the corresponding
+[documentation page](@/documentation/operations/upgrading.md) first, as well as
+the documentation relevant to your version of Garage in the case of major
+upgrades.  With the containerized setup proposed here, the upgrade process
+will require stopping and removing the existing container, and re-creating it
+with the upgraded version.

-## Controling the daemon
+## Controlling the daemon

 The `garage` binary has two purposes:
  - it acts as a daemon when launched with `garage server`
@@ -246,7 +259,7 @@ You can then instruct nodes to connect to one another as follows:
 Venus$ garage node connect 563e1ac825ee3323aa441e72c26d1030d6d4414aeb3dd25287c531e7fc2bc95d@[fc00:1::1]:3901
 ```

-You don't nead to instruct all node to connect to all other nodes:
+You don't need to instruct all node to connect to all other nodes:
 nodes will discover one another transitively.

 Now if your run `garage status` on any node, you should have an output that looks as follows:
@@ -270,12 +283,12 @@ of a role that is assigned to each active cluster node.
 For our example, we will suppose we have the following infrastructure
 (Capacity, Identifier and Zone are specific values to Garage described in the following):

-| Location | Name    | Disk Space | `Capacity` | `Identifier` | `Zone` |
-|----------|---------|------------|------------|--------------|--------------|
-| Paris    | Mercury | 1 TB       | `10`       | `563e`     | `par1`       |
-| Paris    | Venus   | 2 TB       | `20`       | `86f0`     | `par1`       |
-| London   | Earth   | 2 TB       | `20`       | `6814`     | `lon1`       |
-| Brussels | Mars    | 1.5 TB     | `15`       | `212f`     | `bru1`       |
+| Location | Name    | Disk Space | Identifier | Zone (`-z`) | Capacity (`-c`) |
+|----------|---------|------------|------------|-------------|-----------------|
+| Paris    | Mercury | 1 TB       | `563e`     | `par1`      | `1T`            |
+| Paris    | Venus   | 2 TB       | `86f0`     | `par1`      | `2T`            |
+| London   | Earth   | 2 TB       | `6814`     | `lon1`      | `2T`            |
+| Brussels | Mars    | 1.5 TB     | `212f`     | `bru1`      | `1.5T`          |

 #### Node identifiers

@@ -297,6 +310,8 @@ garage status
 It will display the IP address associated with each node;
 from the IP address you will be able to recognize the node.

+We will now use the `garage layout assign` command to configure the correct parameters for each node.
+
 #### Zones

 Zones are simply a user-chosen identifier that identify a group of server that are grouped together logically.
@@ -306,29 +321,29 @@ In most cases, a zone will correspond to a geographical location (i.e. a datacen
 Behind the scene, Garage will use zone definition to try to store the same data on different zones,
 in order to provide high availability despite failure of a zone.

+Zones are passed to Garage using the `-z` flag of `garage layout assign` (see below).
+
 #### Capacity

-Garage reasons on an abstract metric about disk storage that is named the *capacity* of a node.
-The capacity configured in Garage must be proportional to the disk space dedicated to the node.
+Garage needs to know the storage capacity (disk space) it can/should use on
+each node, to be able to correctly balance data.

-Capacity values must be **integers** but can be given any signification.
-Here we chose that 1 unit of capacity = 100 GB.
+Capacity values are expressed in bytes and are passed to Garage using the `-c` flag of `garage layout assign` (see below).

-Note that the amount of data stored by Garage on each server may not be strictly proportional to
-its capacity value, as Garage will priorize having 3 copies of data in different zones,
-even if this means that capacities will not be strictly respected. For example in our above examples,
-nodes Earth and Mars will always store a copy of everything each, and the third copy will
-have 66% chance of being stored by Venus and 33% chance of being stored by Mercury.
+#### Tags
+
+You can add additional tags to nodes using the `-t` flag of `garage layout assign` (see below).
+Tags have no specific meaning for Garage and can be used at your convenience.

 #### Injecting the topology

 Given the information above, we will configure our cluster as follow:

 ```bash
-garage layout assign 563e -z par1 -c 10 -t mercury
-garage layout assign 86f0 -z par1 -c 20 -t venus
-garage layout assign 6814 -z lon1 -c 20 -t earth 
-garage layout assign 212f -z bru1 -c 15 -t mars 
+garage layout assign 563e -z par1 -c 1T -t mercury
+garage layout assign 86f0 -z par1 -c 2T -t venus
+garage layout assign 6814 -z lon1 -c 2T -t earth
+garage layout assign 212f -z bru1 -c 1.5T -t mars
 ```

 At this point, the changes in the cluster layout have not yet been applied.
@@ -338,6 +353,7 @@ To show the new layout that will be applied, call:
 garage layout show
 ```

+Make sure to read carefully the output of `garage layout show`.
 Once you are satisfied with your new layout, apply it with:

 ```bash
@@ -7,7 +7,7 @@ The main reason to add a reverse proxy in front of Garage is to provide TLS to y

 In production you will likely need your certificates signed by a certificate authority.
 The most automated way is to use a provider supporting the [ACME protocol](https://datatracker.ietf.org/doc/html/rfc8555) 
-such as [Let's Encrypt](https://letsencrypt.org/), [ZeroSSL](https://zerossl.com/) or [Buypass Go SSL](https://www.buypass.com/ssl/products/acme).
+such as [Let's Encrypt](https://letsencrypt.org/) or [ZeroSSL](https://zerossl.com/).

 If you are only testing Garage, you can generate a self-signed certificate to follow the documentation:

@@ -97,7 +97,7 @@ server {
  location / {
    proxy_pass http://s3_backend;
    proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-    proxy_set_header Host $host;
+    proxy_set_header Host $http_host;
    # Disable buffering to a temporary file.
    proxy_max_temp_file_size 0;
  }
@@ -142,7 +142,74 @@ server {

 ## Apache httpd

-@TODO
+The [Apache HTTP Server](https://httpd.apache.org/)
+is a general purpose web server that includes
+[reverse proxy](https://httpd.apache.org/docs/2.4/mod/mod_proxy.html)
+capabilities.
+
+### Exposing the S3 endpoints
+
+Create a new [virtual host](https://httpd.apache.org/docs/2.4/vhosts/),
+obtain a certificate using
+[certbot](https://eff-certbot.readthedocs.io/en/stable/using.html#apache),
+and add the
+[`ProxyPass`](https://httpd.apache.org/docs/2.4/mod/mod_proxy.html#proxypass)
+and
+[`ProxyPreserveHost`](https://httpd.apache.org/docs/2.4/mod/mod_proxy.html#proxypreservehost)
+options:
+
+```apache
+<VirtualHost *:443>
+  ServerName garage.example.com
+
+  SSLCertificateFile /etc/letsencrypt/live/garage.example.com/fullchain.pem
+  SSLCertificateKeyFile /etc/letsencrypt/live/garage.example.com/privkey.pem
+  Include /etc/letsencrypt/options-ssl-apache.conf
+
+  Header always set Strict-Transport-Security "max-age=31536000"
+  Header always add Content-Security-Policy upgrade-insecure-requests
+
+  ProxyPass "/" "http://localhost:3900/" nocanon
+  ProxyPreserveHost on
+</VirtualHost>
+```
+
+The `nocanon` keyword is important for
+[presigned URLs](https://docs.aws.amazon.com/AmazonS3/latest/userguide/using-presigned-url.html);
+otherwise,
+
+> `mod_proxy` will canonicalise ProxyPassed URLs.
+> But this may be incompatible with some backends,
+> particularly those that make use of `PATH_INFO`.
+> The optional `nocanon` keyword suppresses this
+> and passes the URL path "raw" to the backend.
+
+### Exposing the web endpoint
+
+Adding static websites backed by Garage works very similarly,
+with the only difference being the port selected in the `ProxyPass` directive.
+
+```apache
+  ProxyPass "/" "http://localhost:3902/" nocanon
+```
+
+### Using Unix sockets
+
+Apache can also proxy via Unix sockets instead of TCP ports,
+if Garage is so configured.
+
+`garage.toml`:
+
+```toml
+[s3_api]
+api_bind_addr = "/run/garage/s3_api.socket"
+```
+
+Apache config:
+
+```apache
+  ProxyPass "/" "unix:/run/garage/s3_api.socket|http://localhost/" nocanon
+```

 ## Traefik v2

@@ -272,7 +339,7 @@ Add the following configuration section [to compress response](https://doc.traef

 ### Add caching response

-Traefik's caching middleware is only available on [entreprise version](https://doc.traefik.io/traefik-enterprise/middlewares/http-cache/), however the freely-available [Souin plugin](https://github.com/darkweak/souin#tr%C3%A6fik-container) can also do the job. (section to be completed)
+Traefik's caching middleware is only available on [enterprise version](https://doc.traefik.io/traefik-enterprise/middlewares/http-cache/), however the freely-available [Souin plugin](https://github.com/darkweak/souin#tr%C3%A6fik-container) can also do the job. (section to be completed)

 ### Complete example

@@ -472,3 +539,32 @@ https:// {

 More information on how this endpoint is implemented in Garage is available
 in the [Admin API Reference](@/documentation/reference-manual/admin-api.md) page.
+
+### Fileserver browser
+
+Caddy's built-in
+[file_server](https://caddyserver.com/docs/caddyfile/directives/file_server)
+browser functionality can be extended with the
+[caddy-fs-s3](https://github.com/sagikazarmark/caddy-fs-s3) module.
+
+This can be configured to use Garage as a backend with the following
+configuration:
+
+```caddy
+browse.garage.tld {
+    file_server {
+        fs s3 {
+            bucket test-bucket
+            region garage
+
+            endpoint https://s3.garage.tld
+            use_path_style
+        }
+
+        browse
+    }
+}
+```
+
+Caddy must also be configured with the required `AWS_ACCESS_KEY_ID` and
+`AWS_SECRET_ACCESS_KEY` environment variables to access the bucket.
@@ -28,6 +28,7 @@ StateDirectory=garage
 DynamicUser=true
 ProtectHome=true
 NoNewPrivileges=true
+LimitNOFILE=42000

 [Install]
 WantedBy=multi-user.target
@@ -37,7 +38,7 @@ WantedBy=multi-user.target
 id is dynamically allocated by systemd (set with `DynamicUser=true`). It cannot
 access (read or write) home folders (`/home`, `/root` and `/run/user`), the
 rest of the filesystem can only be read but not written, only the path seen as
-`/var/lib/garage` is writable as seen by the service. Additionnaly, the process
+`/var/lib/garage` is writable as seen by the service. Additionally, the process
 can not gain new privileges over time.

 For this to work correctly, your `garage.toml` must be set with
@@ -10,7 +10,7 @@ perspective.  It will allow you to understand if Garage is a good fit for
 you, how to better use it, how to contribute to it, what can Garage could
 and could not do, etc.

- **[Goals and use cases](@/documentation/design/goals.md):** This page explains why Garage was concieved and what practical use cases it targets.
+- **[Goals and use cases](@/documentation/design/goals.md):** This page explains why Garage was conceived and what practical use cases it targets.

 - **[Related work](@/documentation/design/related-work.md):** This pages presents the theoretical background on which Garage is built, and describes other software storage solutions and why they didn't work for us.

@@ -31,5 +31,3 @@ We love to talk and hear about Garage, that's why we keep a log here:
  - [(en, 2021-04-28) Distributed object storage is centralised](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/commit/b1f60579a13d3c5eba7f74b1775c84639ea9b51a/doc/talks/2021-04-28_spirals-team/talk.pdf)

  - [(fr, 2020-12-02) Garage : jouer dans la cour des grands quand on est un hébergeur associatif](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/commit/b1f60579a13d3c5eba7f74b1775c84639ea9b51a/doc/talks/2020-12-02_wide-team/talk.pdf)
-
-
@@ -15,14 +15,14 @@ The more a user request will require intra-cluster requests to complete, the mor
 This is especially true for sequential requests: requests that must wait the result of another request to be sent.
 We designed Garage without consensus algorithms (eg. Paxos or Raft) to minimize the number of sequential and parallel requests.

-This serie of benchmarks quantifies the impact of this design choice.
+This series of benchmarks quantifies the impact of this design choice.

 ### On a simple simulated network

 We start with a controlled environment, all the instances are running on the same (powerful enough) machine.

-To control the network latency, we simulate the network with [mknet](https://git.deuxfleurs.fr/trinity-1686a/mknet) (a tool we developped, based on `tc` and the linux network stack).
-To mesure S3 endpoints latency, we use our own tool [s3lat](https://git.deuxfleurs.fr/quentin/s3lat/) to observe only the intra-cluster latency and not some contention on the nodes (CPU, RAM, disk I/O, network bandwidth, etc.).
+To control the network latency, we simulate the network with [mknet](https://git.deuxfleurs.fr/trinity-1686a/mknet) (a tool we developed, based on `tc` and the linux network stack).
+To measure S3 endpoints latency, we use our own tool [s3lat](https://git.deuxfleurs.fr/quentin/s3lat/) to observe only the intra-cluster latency and not some contention on the nodes (CPU, RAM, disk I/O, network bandwidth, etc.).
 Compared to other benchmark tools, S3Lat sends only one (small) request at the same time and measures its latency.
 We selected 5 standard endpoints that are often in the critical path: ListBuckets, ListObjects, GetObject, PutObject and RemoveObject.

@@ -32,7 +32,7 @@ In this first benchmark, we consider 5 instances that are located in a different

 Compared to garage, minio latency drastically increases on 3 endpoints: GetObject, PutObject, RemoveObject.

-We suppose that these requests on minio make transactions over Raft, involving 4 sequential requests: 1) sending the message to the leader, 2) having the leader dispatch it to the other nodes, 3) waiting for the confirmation of followers and finally 4) commiting it. With our current configuration, one Raft transaction will take around 400 ms. GetObject seems to correlate to 1 transaction while PutObject and RemoveObject seems to correlate to 2 or 3. Reviewing minio code would be required to confirm this hypothesis.
+We suppose that these requests on minio make transactions over Raft, involving 4 sequential requests: 1) sending the message to the leader, 2) having the leader dispatch it to the other nodes, 3) waiting for the confirmation of followers and finally 4) committing it. With our current configuration, one Raft transaction will take around 400 ms. GetObject seems to correlate to 1 transaction while PutObject and RemoveObject seems to correlate to 2 or 3. Reviewing minio code would be required to confirm this hypothesis.

 Conversely, garage uses an architecture similar to DynamoDB and never require global cluster coordination to answer a request.
 Instead, garage can always contact the right node in charge of the requested data, and can answer in as low as one request in the case of GetObject and PutObject. We also observed that Garage latency, while often lower to minio, is more dispersed: garage is still in beta and has not received any performance optimization yet.
@@ -50,7 +50,7 @@ We plot a similar graph as before:

 This new graph is very similar to the one before, neither minio or garage seems to benefit from this new topology, but they also do not suffer from it.

-Considering garage, this is expected: nodes in the same DC are put in the same zone, and then data are spread on different zones for data resiliency and availaibility.
+Considering garage, this is expected: nodes in the same DC are put in the same zone, and then data are spread on different zones for data resiliency and availability.
 Then, in the default mode, requesting data requires to query at least 2 zones to be sure that we have the most up to date information.
 These requests will involve at least one inter-DC communication.
 In other words, we prioritize data availability and synchronization over raw performances.
@@ -48,7 +48,24 @@ locations. They use Garage themselves for the following tasks:

 - As a backup target using `rclone` and `restic`

- In the Drone continuous integration platform to store task logs
-
 The Deuxfleurs Garage cluster is a multi-site cluster currently composed of
 9 nodes in 3 physical locations.
+
+### Triplebit
+
+[Triplebit](https://www.triplebit.org) is a non-profit hosting provider and
+ISP focused on improving access to privacy-related services. They use
+Garage themselves for the following tasks:
+
+- Hosting of their homepage, [privacyguides.org](https://www.privacyguides.org/), and various other static sites
+
+- As a PowerDNS authoritative zone backend through [Lightning Stream](https://doc.powerdns.com/lightningstream/latest/index.html) and [LMDB](https://doc.powerdns.com/authoritative/backends/lmdb.html)
+
+- As a Mastodon media storage backend for [mstdn.party](https://mstdn.party/) and [mstdn.plus](https://mstdn.plus/)
+
+- As a PeerTube storage backend for [neat.tube](https://neat.tube/)
+
+- As a [Matrix media backend](https://github.com/matrix-org/synapse-s3-storage-provider)
+
+Triplebit's Garage cluster is a multi-site cluster currently composed of
+15 storage nodes in 3 physical locations.
@@ -94,10 +94,10 @@ delete a tombstone, the following condition has to be met:

 - All nodes responsible for storing this entry are aware of the existence of
  the tombstone, i.e. they cannot hold another version of the entry that is
-  superseeded by the tombstone. This ensures that deleting the tombstone is
+  superseded by the tombstone. This ensures that deleting the tombstone is
  safe and that no deleted value will come back in the system.

-Garage makes use of Sled's atomic operations (such as compare-and-swap and
+Garage uses atomic database operations (such as compare-and-swap and
 transactions) to ensure that only tombstones that have been correctly
 propagated to other nodes are ever deleted from the local entry tree.

@@ -141,4 +141,3 @@ rebalance of data, this would have led to the disk utilization to explode
 during the rebalancing, only to shrink again after 24 hours. The 10-minute
 delay is a compromise that gives good security while not having this problem of
 disk space explosion on rebalance.
-
@@ -37,7 +37,7 @@ However, Amazon S3 source code is not open but alternatives were proposed.
 We identified Minio, Pithos, Swift and Ceph.
 Minio/Ceph enforces a total order, so properties similar to a (relaxed) filesystem.
 Swift and Pithos are probably the most similar to AWS S3 with their consistent hashing ring.
-However Pithos is not maintained anymore. More precisely the company that published Pithos version 1 has developped a second version 2 but has not open sourced it.
+However Pithos is not maintained anymore. More precisely the company that published Pithos version 1 has developed a second version 2 but has not open sourced it.
 Some tests conducted by the [ACIDES project](https://acides.org/) have shown that Openstack Swift consumes way more resources (CPU+RAM) that we can afford. Furthermore, people developing Swift have not designed their software for geo-distribution.

 There were many attempts in research too. I am only thinking to [LBFS](https://pdos.csail.mit.edu/papers/lbfs:sosp01/lbfs.pdf) that was used as a basis for Seafile. But none of them have been effectively implemented yet.
@@ -63,11 +63,11 @@ Due to its industry oriented design, Ceph is also far from being *Simple* to ope
 In a certain way, Ceph and MinIO are closer together than they are from Garage or OpenStack Swift.

 **[Pithos](https://github.com/exoscale/pithos):**
-Pithos has been abandonned and should probably not used yet, in the following we explain why we did not pick their design.
+Pithos has been abandoned and should probably not used yet, in the following we explain why we did not pick their design.
 Pithos was relying as a S3 proxy in front of Cassandra (and was working with Scylla DB too).
 From its designers' mouth, storing data in Cassandra has shown its limitations justifying the project abandonment.
 They built a closed-source version 2 that does not store blobs in the database (only metadata) but did not communicate further on it.
-We considered there v2's design but concluded that it does not fit both our *Self-contained & lightweight* and *Simple* properties. It makes the development, the deployment and the operations more complicated while reducing the flexibility.
+We considered their v2's design but concluded that it does not fit both our *Self-contained & lightweight* and *Simple* properties. It makes the development, the deployment and the operations more complicated while reducing the flexibility.

 **[Riak CS](https://docs.riak.com/riak/cs/2.1.1/index.html):**
 *Not written yet*
@@ -36,7 +36,7 @@ sudo killall nix-daemon
 Now you can enter our nix-shell, all the required packages will be downloaded but they will not pollute your environment outside of the shell:

 ```bash
-nix-shell
+nix-shell -A devShell
 ```

 You can use the traditional Rust development workflow:
@@ -65,8 +65,8 @@ nix-build -j $(nproc) --max-jobs auto
 ```

 Our build has multiple parameters you might want to set:
-  - `release` build with release optimisations instead of debug
-  - `target allows` for cross compilation
+  - `release` to build with release optimisations instead of debug
+  - `target` allows for cross compilation
  - `compileMode` can be set to test or bench to build a unit test runner
  - `git_version` to inject the hash to display when running `garage stats`

@@ -80,13 +80,7 @@ nix-build \
  --git_version $(git rev-parse HEAD)
 ```

-*The result is located in `result/bin`. You can pass arguments to cross compile: check `.drone.yml` for examples.*
-
-If you modify a `Cargo.toml` or regenerate any `Cargo.lock`, you must run `cargo2nix`:
-
-```
-cargo2nix -f
-```
+*The result is located in `result/bin`. You can pass arguments to cross compile: check `.woodpecker/release.yml` for examples.*

 Many tools like rclone, `mc` (minio-client), or `aws` (awscliv2) will be available in your environment and will be useful to test Garage.

@@ -124,23 +118,6 @@ cargo fmt    # format the project, run it before any commit!
 cargo clippy # run the linter, run it before any commit!
 ```

-This is specific to our project, but you will need one last tool, `cargo2nix`.
-To install it, run:
-
-```bash
-cargo install --git https://github.com/superboum/cargo2nix --branch main cargo2nix
-```
-
-You must use it every time you modify a `Cargo.toml` or regenerate a `Cargo.lock` file as follow:
-
-```bash
-cargo build  # Rebuild Cargo.lock if needed
-cargo2nix -f
-```
-
-It will output a `Cargo.nix` file which is a specific `Cargo.lock` file dedicated to Nix that is required by our CI
-which means you must include it in your commits.
-
 Later, to use our scripts and integration tests, you might need additional tools.
 These tools are listed at the end of the `shell.nix` package in the `nativeBuildInputs` part.
 It is up to you to find a way to install the ones you need on your computer.
@@ -3,15 +3,6 @@ title = "Miscellaneous notes"
 weight = 20
 +++

-## Quirks about cargo2nix/rust in Nix
-
-If you use submodules in your crate (like `crdt` and `replication` in `garage_table`), you must list them in `default.nix`
-
-The Windows target does not work. it might be solvable through [overrides](https://github.com/cargo2nix/cargo2nix/blob/master/overlay/overrides.nix). Indeed, we pass `x86_64-pc-windows-gnu` but mingw need `x86_64-w64-mingw32`
-
-We have a simple [PR on cargo2nix](https://github.com/cargo2nix/cargo2nix/pull/201) that fixes critical bugs but the project does not seem very active currently. We must use [my patched version of cargo2nix](https://github.com/superboum/cargo2nix) to enable i686 and armv6l compilation. We might need to contribute to cargo2nix in the future.
-
-
 ## Nix

 Nix has no armv7 + musl toolchains but armv7l is backward compatible with armv6l.
@@ -81,12 +72,9 @@ Our cache will be checked.
 - http://www.lpenz.org/articles/nixchannel/index.html


-## Drone
+## Woodpecker

-Do not try to set a build as trusted from the interface or the CLI tool,
-your request would be ignored. Instead, directly edit the database (table `repos`, column `repo_trusted`).
-
-Drone can do parallelism both at the step and the pipeline level. At the step level, parallelism is restricted to the same runner.
+Woodpecker can do parallelism both at the step and the pipeline level. At the step level, parallelism is restricted to the same runner.

 ## Building Docker containers

@@ -99,3 +87,4 @@ We were:
  - Unable to use the kaniko container provided by Google as we can't run arbitrary logic: we need to put our secret in .docker/config.json.

 Finally we chose to build kaniko through nix and use it in a `nix-shell`.
+We then switched to using kaniko from nixpkgs when it was packaged.
@@ -23,7 +23,7 @@ This logic is defined in `nix/build_index.nix`.
 For each commit, we first pass the code to a formatter (rustfmt) and a linter (clippy).
 Then we try to build it in debug mode and run both unit tests and our integration tests.

-Additionnaly, when releasing, our integration tests are run on the release build for amd64 and i686.
+Additionally, when releasing, our integration tests are run on the release build for amd64 and i686.

 ## Generated Artifacts

@@ -32,7 +32,7 @@ We generate the following binary artifacts for now:
  - **os**: linux
  - **format**: static binary, docker container

-Additionnaly we also build two web pages and one JSON document:
+Additionally we also build two web pages and one JSON document:
  - the documentation (this website)
  - [the release page](https://garagehq.deuxfleurs.fr/_releases.html)
  - [the release list in JSON format](https://garagehq.deuxfleurs.fr/_releases.json)
@@ -42,7 +42,7 @@ and the docker containers on Docker Hub.

 ## Automation

-We automated our release process with Nix and Drone to make it more reliable.
+We automated our release process with Nix and Woodpecker to make it more reliable.
 Here we describe how we have done in case you want to debug or improve it.

 ### Caching build steps
@@ -62,52 +62,31 @@ Sending to the cache is done through `nix copy`, for example:
 nix copy --to 's3://nix?endpoint=garage.deuxfleurs.fr&region=garage&secret-key=/etc/nix/signing-key.sec' result
 ```

-*Note that you need the signing key. In our case, it is stored as a secret in Drone.*
+*The signing key possessed by the Garage maintainers is required to update the Nix cache.*

-The previous command will only send the built packet and not its dependencies.
-To send its dependency, a tool named `nix-copy-closure` has been created but it is not compatible with the S3 protocol.
-
-Instead, you can use the following commands to list all the runtime dependencies:
+The previous command will only send the built package and not its dependencies.
+In the case of our CI pipeline, we want to cache all intermediate build steps
+as well. This can be done using this quite involved command (here as an example
+for the `pkgs.amd64.release` package):

 ```bash
-nix copy \
-  --to 's3://nix?endpoint=garage.deuxfleurs.fr&region=garage&secret-key=/etc/nix/signing-key.sec' \
-  $(nix-store -qR result/)
+nix copy -j8 \
+    --to 's3://nix?endpoint=garage.deuxfleurs.fr&region=garage&secret-key=/etc/nix/nix-signing-key.sec' \
+    $(nix path-info pkgs.amd64.release --file default.nix --derivation --recursive | sed 's/\.drv$/.drv^*/')
 ```

-*We could also write this expression with xargs but this tool is not available in our container.*
+This command will simultaneously build all of the required Nix paths (using at
+most 8 parallel Nix builder jobs) and send the resulting objects to the cache.

-But in certain cases, we want to cache compile time dependencies also.
-For example, the Nix project does not provide binaries for cross compiling to i686 and thus we need to compile gcc on our own.
-We do not want to compile gcc each time, so even if it is a compile time dependency, we want to cache it.
-
-This time, the command is a bit more involved:
-
-```bash
-nix copy --to \
-  's3://nix?endpoint=garage.deuxfleurs.fr&region=garage&secret-key=/etc/nix/signing-key.sec' \
-   $(nix-store -qR --include-outputs \
-     $(nix-instantiate))
-```
-
-This is the command we use in our CI as we expect the final binary to change, so we mainly focus on
-caching our development dependencies.
-
-*Currently there is no automatic garbage collection of the cache: we should monitor its growth.
-Hopefully, we can erase it totally without breaking any build, the next build will only be slower.*
-
-In practise, we concluded that we do not want to cache all the compilation dependencies.
-Instead, we want to cache the toolchain we use to build Garage each time we change it.
-So we removed from Drone any automatic update of the cache and instead handle them manually with:
+This can be run for all the Garage packages we build using the following command:

 ```
 source ~/.awsrc
-nix-shell --run 'refresh_toolchain'
+nix-shell --attr cache --run 'refresh_cache'
 ```

-Internally, it will run `nix-build` on  `nix/toolchain.nix` and send the output plus its depedencies to the cache.
-
-To erase the cache:
+We don't automate this step at each CI build, as *there is currently no automatic garbage collection of the cache.*
+This means we should also monitor the cache's size; if it ever becomes too big we can erase it with:

 ```
 mc rm --recursive --force 'garage/nix/'
@@ -157,9 +136,9 @@ nix-shell --run refresh_index

 If you want to compile for different architectures, you will need to repeat all these commands for each architecture.

-**In practise, and except for debugging, you will never directly run these commands. Release is handled by drone**
+**In practice, and except for debugging, you will never directly run these commands. Release is handled by Woodpecker.**

-### Drone
+### Drone (obsolete)

 Our instance is available at [https://drone.deuxfleurs.fr](https://drone.deuxfleurs.fr).  
 You need an account on [https://git.deuxfleurs.fr](https://git.deuxfleurs.fr) to use it.
@@ -195,5 +174,3 @@ drone sign --save Deuxfleurs/garage
 ```

 Looking at the file, you will see that most of the commands are `nix-shell` and `nix-build` commands with various parameters.
-
-
@@ -19,7 +19,7 @@ connecting to. To run on all nodes, add the `-a` flag as follows:

 # Data block operations

-## Data store scrub
+## Data store scrub {#scrub}

 Scrubbing the data store means examining each individual data block to check that
 their content is correct, by verifying their hash. Any block found to be corrupted
@@ -42,14 +42,14 @@ You may pause an ongoing scrub using `garage repair scrub pause`, but note that
 the scrub will resume automatically 24 hours later as Garage will not let your
 cluster run without a regular scrub. If the scrub procedure is too intensive
 for your servers and is slowing down your workload, the recommended solution
-is to increase the "scrub tranquility" using `garage repair scrub set-tranquility`.
+is to increase the "scrub tranquility" using `garage worker set scrub-tranquility`.
 A higher tranquility value will make Garage take longer pauses between two block
 verifications. Of course, scrubbing the entire data store will also take longer.

 ## Block check and resync

 In some cases, nodes hold a reference to a block but do not actually have the block
-stored on disk. Conversely, they may also have on disk blocks that are not referenced
+stored on disk. Conversely, they may also have on-disk blocks that are not referenced
 any more. To fix both cases, a block repair may be run with `garage repair blocks`.
 This will scan the entire block reference counter table to check that the blocks
 exist on disk, and will scan the entire disk store to check that stored blocks
@@ -91,9 +91,37 @@ is definitely lost, then there is no other choice than to declare your S3 object
 as unrecoverable, and to delete them properly from the data store. This can be done
 using the `garage block purge` command.

+## Rebalancing data directories
+
+In [multi-HDD setups](@/documentation/operations/multi-hdd.md), to ensure that
+data blocks are well balanced between storage locations, you may run a
+rebalance operation using `garage repair rebalance`. This is useful when
+adding storage locations or when capacities of the storage locations have been
+changed.  Once this is finished, Garage will know for each block of a single
+possible location where it can be, which can increase access speed.  This
+operation will also move out all data from locations marked as read-only.
+

 # Metadata operations

+## Metadata snapshotting
+
+It is good practice to setup automatic snapshotting of your metadata database
+file, to recover from situations where it becomes corrupted on disk. This can
+be done at the filesystem level if you are using ZFS or BTRFS.
+
+Since Garage v0.9.4, Garage is able to take snapshots of the metadata database
+itself. This basically amounts to copying the database file, except that it can
+be run live while Garage is running without the risk of corruption or
+inconsistencies.  This can be setup to run automatically on a schedule using
+[`metadata_auto_snapshot_interval`](@/documentation/reference-manual/configuration.md#metadata_auto_snapshot_interval).
+A snapshot can also be triggered manually using the `garage meta snapshot`
+command. Note that taking a snapshot using this method is very intensive as it
+requires making a full copy of the database file, so you might prefer using
+filesystem-level snapshots if possible. To recover a corrupted node from such a
+snapshot, read the instructions
+[here](@/documentation/operations/recovering.md#corrupted_meta).
+
 ## Metadata table resync

 Garage automatically resyncs all entries stored in the metadata tables every hour,
@@ -113,5 +141,7 @@ blocks may still be held by Garage. If you suspect that such corruption has occu
 in your cluster, you can run one of the following repair procedures:

 - `garage repair versions`: checks that all versions belong to a non-deleted object, and purges any orphan version
- `garage repair block_refs`: checks that all block references belong to a non-deleted object version, and purges any orphan block reference (this will then allow the blocks to be garbage-collected)

+- `garage repair block-refs`: checks that all block references belong to a non-deleted object version, and purges any orphan block reference (this will then allow the blocks to be garbage-collected)
+
+- `garage repair block-rc`: checks that the reference counters for blocks are in sync with the actual number of non-deleted entries in the block reference table
@@ -9,18 +9,30 @@ a certain capacity, or a gateway node that does not store data and is only
 used as an API entry point for faster cluster access.
 An introduction to building cluster layouts can be found in the [production deployment](@/documentation/cookbook/real-world.md) page.

+In Garage, all of the data that can be stored in a given cluster is divided
+into slices which we call *partitions*. Each partition is stored by
+one or several nodes in the cluster
+(see [`replication_factor`](@/documentation/reference-manual/configuration.md#replication_factor)).
+The layout determines the correspondence between these partitions,
+which exist on a logical level, and actual storage nodes.
+
 ## How cluster layouts work in Garage

-In Garage, a cluster layout is composed of the following components:
+A cluster layout is composed of the following components:

- a table of roles assigned to nodes
+- a table of roles assigned to nodes, defined by the user
+- an optimal assignation of partitions to nodes, computed by an algorithm that is ran once when calling `garage layout apply` or the ApplyClusterLayout API endpoint
 - a version number

 Garage nodes will always use the cluster layout with the highest version number.

 Garage nodes also maintain and synchronize between them a set of proposed role
 changes that haven't yet been applied. These changes will be applied (or
-canceled) in the next version of the layout
+canceled) in the next version of the layout.
+
+All operations on the layout can be realized using the `garage` CLI or using the
+[administration API endpoint](@/documentation/reference-manual/admin-api.md).
+We give here a description of CLI commands, the admin API semantics are very similar.

 The following commands insert modifications to the set of proposed role changes
 for the next layout version (but they do not create the new layout immediately):
@@ -51,7 +63,7 @@ commands will fail otherwise.

 ## Warnings about Garage cluster layout management

-**Warning: never make several calls to `garage layout apply` or `garage layout
+**⚠️ Never make several calls to `garage layout apply` or `garage layout
 revert` with the same value of the `--version` flag. Doing so can lead to the
 creation of several different layouts with the same version number, in which
 case your Garage cluster will become inconsistent until fixed.** If a call to
@@ -65,13 +77,198 @@ shell, you shouldn't have much issues as long as you run commands one after
 the other and take care of checking the output of `garage layout show`
 before applying any changes.

-If you are using the `garage` CLI to script layout changes, follow the following recommendations:
+If you are using the `garage` CLI or the admin API to script layout changes,
+follow the following recommendations:

- Make all of your `garage` CLI calls to the same RPC host. Do not use the
-  `garage` CLI to connect to individual nodes to send them each a piece of the
-  layout changes you are making, as the changes propagate asynchronously
-  between nodes and might not all be taken into account at the time when the
-  new layout is applied.
+- If using the CLI, make all of your `garage` CLI calls to the same RPC host.
+  If using the admin API, make all of your API calls to the same Garage node. Do
+  not connect to individual nodes to send them each a piece of the layout changes
+  you are making, as the changes propagate asynchronously between nodes and might
+  not all be taken into account at the time when the new layout is applied.

- **Only call `garage layout apply` once**, and call it **strictly after** all
-  of the `layout assign` and `layout remove` commands have returned.
+- **Only call `garage layout apply`/ApplyClusterLayout once**, and call it
+  **strictly after** all of the `layout assign` and `layout remove`
+  commands/UpdateClusterLayout API calls have returned.
+
+
+## Understanding unexpected layout calculations
+
+When adding, removing or modifying nodes in a cluster layout, sometimes
+unexpected assignations of partitions to node can occur. These assignations
+are in fact normal and logical, given the objectives of the algorithm.  Indeed,
+**the layout algorithm prioritizes moving less data between nodes over
+achieving equal distribution of load. It also tries to use all links between
+pairs of nodes in equal proportions when moving data.**  This section presents
+two examples and illustrates how one can control Garage's behavior to obtain
+the desired results.
+
+### Example 1
+
+In this example, a cluster is originally composed of 3 nodes in 3 different
+zones (data centers).  The three nodes are of equal capacity, therefore they
+are all fully exploited and all store a copy of all of the data in the cluster.
+
+Then, a fourth node of the same size is added in the datacenter `dc1`.
+As illustrated by the following, **Garage will by default not store any data on the new node**:
+
+```
+$ garage layout show
+==== CURRENT CLUSTER LAYOUT ====
+ID                Tags   Zone  Capacity   Usable capacity
+b10c110e4e854e5a  node1  dc1   1000.0 MB  1000.0 MB (100.0%)
+a235ac7695e0c54d  node2  dc2   1000.0 MB  1000.0 MB (100.0%)
+62b218d848e86a64  node3  dc3   1000.0 MB  1000.0 MB (100.0%)
+
+Zone redundancy: maximum
+
+Current cluster layout version: 6
+
+==== STAGED ROLE CHANGES ====
+ID                Tags   Zone  Capacity
+a11c7cf18af29737  node4  dc1   1000.0 MB
+
+
+==== NEW CLUSTER LAYOUT AFTER APPLYING CHANGES ====
+ID                Tags   Zone  Capacity   Usable capacity
+b10c110e4e854e5a  node1  dc1   1000.0 MB  1000.0 MB (100.0%)
+a11c7cf18af29737  node4  dc1   1000.0 MB  0 B (0.0%)
+a235ac7695e0c54d  node2  dc2   1000.0 MB  1000.0 MB (100.0%)
+62b218d848e86a64  node3  dc3   1000.0 MB  1000.0 MB (100.0%)
+
+Zone redundancy: maximum
+
+==== COMPUTATION OF A NEW PARTITION ASSIGNATION ====
+
+Partitions are replicated 3 times on at least 3 distinct zones.
+
+Optimal partition size:                     3.9 MB (3.9 MB in previous layout)
+Usable capacity / total cluster capacity:   3.0 GB / 4.0 GB (75.0 %)
+Effective capacity (replication factor 3):  1000.0 MB
+
+A total of 0 new copies of partitions need to be transferred.
+
+dc1                 Tags   Partitions        Capacity   Usable capacity
+  b10c110e4e854e5a  node1  256 (0 new)       1000.0 MB  1000.0 MB (100.0%)
+  a11c7cf18af29737  node4  0 (0 new)         1000.0 MB  0 B (0.0%)
+  TOTAL                    256 (256 unique)  2.0 GB     1000.0 MB (50.0%)
+
+dc2                 Tags   Partitions        Capacity   Usable capacity
+  a235ac7695e0c54d  node2  256 (0 new)       1000.0 MB  1000.0 MB (100.0%)
+  TOTAL                    256 (256 unique)  1000.0 MB  1000.0 MB (100.0%)
+
+dc3                 Tags   Partitions        Capacity   Usable capacity
+  62b218d848e86a64  node3  256 (0 new)       1000.0 MB  1000.0 MB (100.0%)
+  TOTAL                    256 (256 unique)  1000.0 MB  1000.0 MB (100.0%)
+```
+
+While unexpected, this is logical because of the following facts:
+
+- storing some data on the new node does not help increase the total quantity
+  of data that can be stored on the cluster, as the two other zones (`dc2` and
+  `dc3`) still need to store a full copy of everything, and their capacity is
+  still the same;
+
+- there is therefore no need to move any data on the new node as this would be pointless;
+
+- moving data to the new node has a cost which the algorithm decides to not pay if not necessary.
+
+This distribution of data can however not be what the administrator wanted: if
+they added a new node to `dc1`, it might be because the existing node is too
+slow, and they wish to divide its load by half. In that case, what they need to
+do to force Garage to distribute the data between the two nodes is to attribute
+only half of the capacity to each node in `dc1` (in our example, 500M instead of 1G).
+In that case, Garage would determine that to be able to store 1G in total, it
+would need to store 500M on the old node and 500M on the added one.
+
+
+### Example 2
+
+The following example is a slightly different scenario, where `dc1` had two
+nodes that were used at 50%, and `dc2` and `dc3` each have one node that is
+100% used. All node capacities are the same.
+
+Then, a node from `dc1` is moved into `dc3`. One could expect that the roles of
+`dc1` and `dc3` would simply be swapped: the remaining node in `dc1` would be
+used at 100%, and the two nodes now in `dc3` would be used at 50%. Instead,
+this happens:
+
+```
+==== CURRENT CLUSTER LAYOUT ====
+ID                Tags   Zone  Capacity   Usable capacity
+b10c110e4e854e5a  node1  dc1   1000.0 MB  500.0 MB (50.0%)
+a11c7cf18af29737  node4  dc1   1000.0 MB  500.0 MB (50.0%)
+a235ac7695e0c54d  node2  dc2   1000.0 MB  1000.0 MB (100.0%)
+62b218d848e86a64  node3  dc3   1000.0 MB  1000.0 MB (100.0%)
+
+Zone redundancy: maximum
+
+Current cluster layout version: 8
+
+==== STAGED ROLE CHANGES ====
+ID                Tags   Zone  Capacity
+a11c7cf18af29737  node4  dc3   1000.0 MB
+
+
+==== NEW CLUSTER LAYOUT AFTER APPLYING CHANGES ====
+ID                Tags   Zone  Capacity   Usable capacity
+b10c110e4e854e5a  node1  dc1   1000.0 MB  1000.0 MB (100.0%)
+a235ac7695e0c54d  node2  dc2   1000.0 MB  1000.0 MB (100.0%)
+62b218d848e86a64  node3  dc3   1000.0 MB  753.9 MB (75.4%)
+a11c7cf18af29737  node4  dc3   1000.0 MB  246.1 MB (24.6%)
+
+Zone redundancy: maximum
+
+==== COMPUTATION OF A NEW PARTITION ASSIGNATION ====
+
+Partitions are replicated 3 times on at least 3 distinct zones.
+
+Optimal partition size:                     3.9 MB (3.9 MB in previous layout)
+Usable capacity / total cluster capacity:   3.0 GB / 4.0 GB (75.0 %)
+Effective capacity (replication factor 3):  1000.0 MB
+
+A total of 128 new copies of partitions need to be transferred.
+
+dc1                 Tags   Partitions        Capacity   Usable capacity
+  b10c110e4e854e5a  node1  256 (128 new)     1000.0 MB  1000.0 MB (100.0%)
+  TOTAL                    256 (256 unique)  1000.0 MB  1000.0 MB (100.0%)
+
+dc2                 Tags   Partitions        Capacity   Usable capacity
+  a235ac7695e0c54d  node2  256 (0 new)       1000.0 MB  1000.0 MB (100.0%)
+  TOTAL                    256 (256 unique)  1000.0 MB  1000.0 MB (100.0%)
+
+dc3                 Tags   Partitions        Capacity   Usable capacity
+  62b218d848e86a64  node3  193 (0 new)       1000.0 MB  753.9 MB (75.4%)
+  a11c7cf18af29737  node4  63 (0 new)        1000.0 MB  246.1 MB (24.6%)
+  TOTAL                    256 (256 unique)  2.0 GB     1000.0 MB (50.0%)
+```
+
+As we can see, the node that was moved to `dc3` (node4) is only used at 25% (approximately),
+whereas the node that was already in `dc3` (node3) is used at 75%.
+
+This can be explained by the following:
+
+- node1 will now be the only node remaining in `dc1`, thus it has to store all
+  of the data in the cluster. Since it was storing only half of it before, it has
+  to retrieve the other half from other nodes in the cluster.
+
+- The data which it does not have is entirely stored by the other node that was
+  in `dc1` and that is now in `dc3` (node4). There is also a copy of it on node2
+  and node3 since both these nodes have a copy of everything.
+
+- node3 and node4 are the two nodes that will now be in a datacenter that is
+ under-utilized (`dc3`), this means that those are the two candidates from which
+ data can be removed to be moved to node1.
+
+- Garage will move data in equal proportions from all possible sources, in this
+  case it means that it will transfer 25% of the entire data set from node3 to
+  node1 and another 25% from node4 to node1.
+
+This explains why node3 ends with 75% utilization (100% from before minus 25%
+that is moved to node1), and node4 ends with 25% (50% from before minus 25%
+that is moved to node1).
+
+This illustrates the second principle of the layout computation: **if there is
+a choice in moving data out of some nodes, then all links between pairs of
+nodes are used in equal proportions** (this is approximately true, there is
+randomness in the algorithm to achieve this so there might be some small
+fluctuations, as we see above).
@@ -0,0 +1,101 @@
+++
+title = "Multi-HDD support"
+weight = 15
+++
+
+
+Since v0.9, Garage natively supports nodes that have several storage drives
+for storing data blocks (not for metadata storage).
+
+## Initial setup
+
+To set up a new Garage storage node with multiple HDDs,
+format and mount all your drives in different directories,
+and use a Garage configuration as follows:
+
+```toml
+data_dir = [
+    { path = "/path/to/hdd1", capacity = "2T" },
+    { path = "/path/to/hdd2", capacity = "4T" },
+]
+```
+
+Garage will automatically balance all blocks stored by the node
+among the different specified directories, proportionally to the
+specified capacities.
+
+## Updating the list of storage locations
+
+If you add new storage locations to your `data_dir`,
+Garage will not rebalance existing data between storage locations.
+Newly written blocks will be balanced proportionally to the specified capacities,
+and existing data may be moved between drives to improve balancing,
+but only opportunistically when a data block is re-written (e.g. an object
+is re-uploaded, or an object with a duplicate block is uploaded).
+
+To understand precisely what is happening, we need to dive in to how Garage
+splits data among the different storage locations.
+
+First of all, Garage divides the set of all possible block hashes
+in a fixed number of slices (currently 1024), and assigns
+to each slice a primary storage location among the specified data directories.
+The number of slices having their primary location in each data directory
+is proportional to the capacity specified in the config file.
+
+When Garage receives a block to write, it will always write it in the primary
+directory of the slice that contains its hash.
+
+Now, to be able to not lose existing data blocks when storage locations
+are added, Garage also keeps a list of secondary data directories
+for all of the hash slices. Secondary data directories for a slice indicates
+storage locations that once were primary directories for that slice, i.e. where
+Garage knows that data blocks of that slice might be stored.
+When Garage is requested to read a certain data block,
+it will first look in the primary storage directory of its slice,
+and if it doesn't find it there it goes through all of the secondary storage
+locations until it finds it. This allows Garage to continue operating
+normally when storage locations are added, without having to shuffle
+files between drives to place them in the correct location.
+
+This relatively simple strategy works well but does not ensure that data
+is correctly balanced among drives according to their capacity.
+To rebalance data, two strategies can be used:
+
+- Lazy rebalancing: when a block is re-written (e.g. the object is re-uploaded),
+  Garage checks whether the existing copy is in the primary directory of the slice
+  or in a secondary directory. If the current copy is in a secondary directory,
+  Garage re-writes a copy in the primary directory and deletes the one from the
+  secondary directory. This might never end up rebalancing everything if there
+  are data blocks that are only read and never written.
+
+- Active rebalancing: an operator of a Garage node can explicitly launch a repair
+  procedure that rebalances the data directories, moving all blocks to their
+  primary location. Once done, all secondary locations for all hash slices are
+  removed so that they won't be checked anymore when looking for a data block.
+
+## Read-only storage locations
+
+If you would like to move all data blocks from an existing data directory to one
+or several new data directories, mark the old directory as read-only:
+
+```toml
+data_dir = [
+    { path = "/path/to/old_data", read_only = true },
+    { path = "/path/to/new_hdd1", capacity = "2T" },
+    { path = "/path/to/new_hdd2", capacity = "4T" },
+]
+```
+
+Garage will be able to read requested blocks from the read-only directory.
+Garage will also move data out of the read-only directory either progressively
+(lazy rebalancing) or if requested explicitly (active rebalancing).
+
+Once an active rebalancing has finished, your read-only directory should be empty:
+it might still contain subdirectories, but no data files. You can check that
+it contains no files using:
+
+```bash
+find -type f /path/to/old_data      # should not print anything
+```
+
+at which point it can be removed from the `data_dir` list in your config file.
@@ -5,7 +5,7 @@ weight = 40

 Garage is meant to work on old, second-hand hardware.
 In particular, this makes it likely that some of your drives will fail, and some manual intervention will be needed.
-Fear not! For Garage is fully equipped to handle drive failures, in most common cases.
+Fear not! Garage is fully equipped to handle drive failures, in most common cases.

 ## A note on availability of Garage

@@ -61,7 +61,7 @@ garage repair -a --yes blocks

 This will re-synchronize blocks of data that are missing to the new HDD, reading them from copies located on other nodes.

-You can check on the advancement of this process by doing the following command: 
+You can check on the advancement of this process by doing the following command:

 ```bash
 garage stats -a
@@ -108,3 +108,60 @@ garage layout apply   # once satisfied, apply the changes

 Garage will then start synchronizing all required data on the new node.
 This process can be monitored using the `garage stats -a` command.
+
+## Replacement scenario 3: corrupted metadata {#corrupted_meta}
+
+In some cases, your metadata DB file might become corrupted, for instance if
+your node suffered a power outage and did not shut down properly. In this case,
+you can recover without having to change the node ID and rebuilding a cluster
+layout. This means that data blocks will not need to be shuffled around, you
+must simply find a way to repair the metadata file. The best way is generally
+to discard the corrupted file and recover it from another source.
+
+First of all, start by locating the database file in your metadata directory,
+which [depends on your `db_engine`
+choice](@/documentation/reference-manual/configuration.md#db_engine).  Then,
+your recovery options are as follows:
+
+- **Option 1: resyncing from other nodes.** In case your cluster is replicated
+  with two or three copies, you can simply delete the database file, and Garage
+  will resync from other nodes. To do so, stop Garage, delete the database file
+  or directory, and restart Garage. Then, do a full table repair by calling
+  `garage repair -a --yes tables`.  This will take a bit of time to complete as
+  the new node will need to receive copies of the metadata tables from the
+  network.
+
+- **Option 2: restoring a snapshot taken by Garage.** Since v0.9.4, Garage can
+  [automatically take regular
+  snapshots](@/documentation/reference-manual/configuration.md#metadata_auto_snapshot_interval)
+  of your metadata DB file. This file or directory should be located under
+  `<metadata_dir>/snapshots`, and is named according to the UTC time at which it
+  was taken. Stop Garage, discard the database file/directory and replace it by the
+  snapshot you want to use. For instance, in the case of LMDB:
+
+  ```bash
+  cd $METADATA_DIR
+  mv db.lmdb db.lmdb.bak
+  cp -r snapshots/2024-03-15T12:13:52Z db.lmdb
+  ```
+
+  And for Sqlite:
+
+  ```bash
+  cd $METADATA_DIR
+  mv db.sqlite db.sqlite.bak
+  cp snapshots/2024-03-15T12:13:52Z db.sqlite
+  ```
+
+  Then, restart Garage and run a full table repair by calling `garage repair -a
+  --yes tables`.  This should run relatively fast as only the changes that
+  occurred since the snapshot was taken will need to be resynchronized. Of
+  course, if your cluster is not replicated, you will lose all changes that
+  occurred since the snapshot was taken.
+
+- **Option 3: restoring a filesystem-level snapshot.** If you are using ZFS or
+  BTRFS to snapshot your metadata partition, refer to their specific
+  documentation on rolling back or copying files from an old snapshot. 
+  Note that, depending on the properties of the filesystem and of the DB engine,
+  if these snapshots were taken during a write operation to the database, they may
+  also be corrupted and thus unfit for recovery.
@@ -9,7 +9,7 @@ On a new version release, there is 2 possibilities:
  - protocols and data structures remained the same ➡️ this is a **minor upgrade**
  - protocols or data structures changed  ➡️  this is a **major upgrade**

-You can quickly now what type of update you will have to operate by looking at the version identifier:
+You can quickly know what type of update you will have to operate by looking at the version identifier:
 when we require our users to do a major upgrade, we will always bump the first nonzero component of the version identifier
 (e.g. from v0.7.2 to v0.8.0).
 Conversely, for versions that only require a minor upgrade, the first nonzero component will always stay the same (e.g. from v0.8.0 to v0.8.1).
@@ -56,7 +56,7 @@ From a high level perspective, a major upgrade looks like this:
  10. Enable API access (reverse step 1)
  11. Monitor your cluster while load comes back, check that all your applications are happy with this new version

-### Major upgarades with minimal downtime
+### Major upgrades with minimal downtime

 There is only one operation that has to be coordinated cluster-wide: the switch of one version of the internal RPC protocol to the next.
 This means that an upgrade with very limited downtime can simply be performed from one major version to the next by restarting all nodes
@@ -71,7 +71,19 @@ The entire procedure would look something like this:

 2. Take each node offline individually to back up its metadata folder, bring them back online once the backup is done.
  You can do all of the nodes in a single zone at once as that won't impact global cluster availability.
-  Do not try to make a backup of the metadata folder of a running node.
+  Do not try to manually copy the metadata folder of a running node.
+
+  **Since Garage v0.9.4,** you can use the `garage meta snapshot --all` command
+  to take a simultaneous snapshot of the metadata database files of all your
+  nodes.  This avoids the tedious process of having to take them down one by
+  one before upgrading. Be careful that if automatic snapshotting is enabled,
+  Garage only keeps the last two snapshots and deletes older ones, so you might
+  want to disable automatic snapshotting in your upgraded configuration file
+  until you have confirmed that the upgrade ran successfully.  In addition to
+  snapshotting the metadata databases of your nodes, you should back-up at
+  least the `cluster_layout` file of one of your Garage instances (this file
+  should be the same on all nodes and you can copy it safely while Garage is
+  running).

 3. Prepare your binaries and configuration files for the new Garage version

@@ -80,6 +92,6 @@ The entire procedure would look something like this:
 5. If any specific migration procedure is required, it is usually in one of the two cases:

  - It can be run on online nodes after the new version has started, during regular cluster operation.
-  - it has to be run offline
+  - it has to be run offline, in which case you will have to again take all nodes offline one after the other to run the repair

   For this last step, please refer to the specific documentation pertaining to the version upgrade you are doing.
@@ -42,6 +42,11 @@ If a binary of the last version is not available for your architecture,
 or if you want a build customized for your system,
 you can [build Garage from source](@/documentation/cookbook/from-source.md).

+If none of these option work for you, you can also run Garage in a Docker
+container.  For simplicity, a minimal command to launch Garage using Docker is
+provided in this quick start guide.  We recommend reading the tutorial on
+[configuring a multi-node cluster](@/documentation/cookbook/real-world.md) to
+learn about the full Docker workflow for Garage.

 ## Configuring and starting Garage

@@ -57,9 +62,9 @@ to generate unique and private secrets for security reasons:
 cat > garage.toml <<EOF
 metadata_dir = "/tmp/meta"
 data_dir = "/tmp/data"
-db_engine = "lmdb"
+db_engine = "sqlite"

-replication_mode = "none"
+replication_factor = 1

 rpc_bind_addr = "[::]:3901"
 rpc_public_addr = "127.0.0.1:3901"
@@ -75,20 +80,23 @@ bind_addr = "[::]:3902"
 root_domain = ".web.garage.localhost"
 index = "index.html"

-[k2v_api]
-api_bind_addr = "[::]:3904"
-
 [admin]
-api_bind_addr = "0.0.0.0:3903"
+api_bind_addr = "[::]:3903"
 admin_token = "$(openssl rand -base64 32)"
+metrics_token = "$(openssl rand -base64 32)"
 EOF
 ```

-Now that your configuration file has been created, you can put
-it in the right place. By default, garage looks at **`/etc/garage.toml`.**
+See the [Configuration file format](https://garagehq.deuxfleurs.fr/documentation/reference-manual/configuration/)
+for complete options and values.

-You can also store it somewhere else, but you will have to specify `-c path/to/garage.toml`
-at each invocation of the `garage` binary (for example: `garage -c ./garage.toml server`, `garage -c ./garage.toml status`).
+By default, Garage looks for its configuration file in **`/etc/garage.toml`.**
+Since we have written our configuration file in the working directory, we will have to set
+the following environment variable:
+
+```bash
+export GARAGE_CONFIG_FILE=$(pwd)/garage.toml
+```

 As you can see, the `rpc_secret` is a 32 bytes hexadecimal string.
 You can regenerate it with `openssl rand -hex 32`.
@@ -101,18 +109,99 @@ Garage server will not be persistent. Change these to locations on your local di
 your data to be persisted properly.


+### Configuring initial access credentials
+
+Since `v2.3.0`, Garage can automatically create a default access key and a default storage bucket,
+based on values provided in environment variables.
+
+To use this feature, export the following environment variables:
+
+```bash
+export GARAGE_DEFAULT_ACCESS_KEY="GK$(openssl rand -hex 16)"
+export GARAGE_DEFAULT_SECRET_KEY="$(openssl rand -hex 32)"
+export GARAGE_DEFAULT_BUCKET="default-bucket"
+```
+
+The example above creates a random access key ID and associated secret key.
+You can also provide an access key ID and secret key of your own.
+
 ### Launching the Garage server

-Use the following command to launch the Garage server with our configuration file:
+Use the following command to launch the Garage server:

-```
-garage server
+```bash
+garage server --single-node --default-bucket
 ```

-You can tune Garage's verbosity as follows (from less verbose to more verbose):
+The `--single-node` flag instructs Garage to automatically configure a single-node cluster without data replication.
+The `--default-bucket` flag instructs Garage to create a default access key and a default bucket using the environment variables we defined above.
+Both flags are optional and can be omitted, in which case you will have to follow manual configuration steps described below.
+
+**For older versions of Garage (before v2.3.0):** automatic configuration using `--single-node` and `--default-bucket` is not available,
+you must follow the manual configuration steps.
+
+Alternatively, if you cannot or do not wish to run the Garage binary directly,
+you may use Docker to run Garage in a container using the following command:
+
+```bash
+docker run \
+  -d \
+  --name garage-container \
+  -p 3900:3900 -p 3901:3901 -p 3902:3902 -p 3903:3903 \
+  -v $(pwd)/garage.toml:/etc/garage.toml \
+  -e GARAGE_DEFAULT_ACCESS_KEY \
+  -e GARAGE_DEFAULT_SECRET_KEY \
+  -e GARAGE_DEFAULT_BUCKET \
+  dxflrs/garage:v2.3.0
+  /garage server --single-node --default-bucket
+```
+
+Note that this command will NOT create persistent volumes for Garage's data, so
+your cluster will be wiped if the container terminates.  To persist Garage's
+data, you must manually add volumes for the `data` and `metadata` directories
+and configure their correct paths in your `garage.toml` files (see [configuring
+a multi-node cluster](@/documentation/cookbook/real-world.md)).
+
+Under Linux, you can substitute `--network host` for `-p 3900:3900 -p 3901:3901 -p 3902:3902 -p 3903:3903`.
+
+### Checking that Garage runs correctly
+
+The `garage` utility is also used as a CLI tool to administrate your Garage
+deployment.  It needs read access to your configuration file and to the metadata directory
+to obtain connection parameters to contact the local Garage node.
+
+Use the following command to show the status of your cluster:

 ```
-RUST_LOG=garage=info garage server
+garage status
+```
+
+If you are running Garage in a Docker container, you can use the following command instead:
+
+```bash
+docker exec garage-container /garage status
+```
+
+This should show something like this:
+
+```
+==== HEALTHY NODES ====
+ID                Hostname  Address         Tags       Zone  Capacity  DataAvail         Version
+563e1ac825ee3323  linuxbox  127.0.0.1:3901  [default]  dc1   19.9 GiB  19.5 GiB (97.6%)  v2.3.0
+```
+
+### Troubleshooting
+
+Ensure your configuration file, `metadata_dir` and `data_dir`  are readable by the user running the `garage` server or Docker.
+
+When running the `garage` CLI, ensure that the path to your configuration file is correctly specified (see below),
+and that it can read it and read from your metadata directory.
+
+You can tune Garage's verbosity by setting the `RUST_LOG=` environment variable.
+Available log levels are (from less verbose to more verbose): `error`, `warn`, `info` *(default)*, `debug` and `trace`.
+
+```bash
+RUST_LOG=garage=info garage server # default
 RUST_LOG=garage=debug garage server
 RUST_LOG=garage=trace garage server
 ```
@@ -121,58 +210,97 @@ Log level `info` is the default value and is recommended for most use cases.
 Log level `debug` can help you check why your S3 API calls are not working.


-### Checking that Garage runs correctly

-The `garage` utility is also used as a CLI tool to configure your Garage deployment.
-It uses values from the TOML configuration file to find the Garage daemon running on the
-local node, therefore if your configuration file is not at `/etc/garage.toml` you will
-again have to specify `-c path/to/garage.toml`.
+## Uploading and downloading from Garage

-If the `garage` CLI is able to correctly detect the parameters of your local Garage node,
-the following command should be enough to show the status of your cluster:
+This section will show how to download and upload files on Garage using a third-party tool named `awscli`.

-```
-garage status
-```

-This should show something like this:
+### Install and configure `awscli`

-```
-==== HEALTHY NODES ====
-ID                 Hostname  Address         Tag                   Zone  Capacity
-563e1ac825ee3323…  linuxbox  127.0.0.1:3901  NO ROLE ASSIGNED
-```
-
-## Creating a cluster layout
-
-Creating a cluster layout for a Garage deployment means informing Garage
-of the disk space available on each node of the cluster
-as well as the zone (e.g. datacenter) each machine is located in.
-
-For our test deployment, we are using only one node. The way in which we configure
-it does not matter, you can simply write:
+If you have python on your system, you can install it with:

 ```bash
-garage layout assign -z dc1 -c 1 <node_id>
+python -m pip install --user awscli
 ```

-where `<node_id>` corresponds to the identifier of the node shown by `garage status` (first column).
-You can enter simply a prefix of that identifier.
-For instance here you could write just `garage layout assign -z dc1 -c 1 563e`.
-
-The layout then has to be applied to the cluster, using:
+Now that `awscli` is installed, you must configure it to talk to your Garage
+instance using the credentials defined above. Here is a simple way to create
+a configuration file in `~/.awsrc` using a single command that will save the
+secrets from your environment:

 ```bash
-garage layout apply
+cat > ~/.awsrc <<EOF
+export AWS_ENDPOINT_URL='http://localhost:3900'
+export AWS_DEFAULT_REGION='garage'
+export AWS_ACCESS_KEY_ID='$GARAGE_DEFAULT_ACCESS_KEY'
+export AWS_SECRET_ACCESS_KEY='$GARAGE_DEFAULT_SECRET_KEY'
+
+aws --version
+EOF
+
 ```

+Note that you need to have at least `awscli` `>=1.29.0` or `>=2.13.0`, otherwise you
+need to specify `--endpoint-url` explicitly on each `awscli` invocation.

-## Creating buckets and keys
+Now, each time you want to use `awscli` on this target, run:

-In this section, we will suppose that we want to create a bucket named `nextcloud-bucket`
-that will be accessed through a key named `nextcloud-app-key`.
+```bash
+source ~/.awsrc
+```

-Don't forget that `help` command and `--help` subcommands can help you anywhere,
+*You can create multiple files with different names if you
+have multiple Garage clusters or different keys.
+Switching from one cluster to another is as simple as
+sourcing the right file.*
+
+### Example usage of `awscli`
+
+```bash
+# list buckets
+aws s3 ls
+
+# list objects of a bucket
+aws s3 ls s3://default-bucket
+
+# copy from your filesystem to garage
+aws s3 cp /proc/cpuinfo s3://default-bucket/cpuinfo.txt
+
+# copy from garage to your filesystem
+aws s3 cp s3://default-bucket/cpuinfo.txt /tmp/cpuinfo.txt
+```
+
+Note that you can use `awscli` for more advanced operations like
+creating a bucket, pre-signing a request or managing your website.
+[Read the full documentation to know more](https://awscli.amazonaws.com/v2/documentation/api/latest/reference/s3/index.html).
+
+Some features are however not implemented like ACL or policy.
+Check [our S3 compatibility list](@/documentation/reference-manual/s3-compatibility.md).
+
+### Other tools for interacting with Garage
+
+The following tools can also be used to send and receive files from/to Garage:
+
+- [minio-client](@/documentation/connect/cli.md#minio-client)
+- [s3cmd](@/documentation/connect/cli.md#s3cmd)
+- [rclone](@/documentation/connect/cli.md#rclone)
+- [Cyberduck](@/documentation/connect/cli.md#cyberduck)
+- [WinSCP](@/documentation/connect/cli.md#winscp)
+
+An exhaustive list is maintained in the ["Integrations" > "Browsing tools" section](@/documentation/connect/_index.md).
+
+
+
+## Manual configuration
+
+This section provides instructions that are equivalent to using the
+`--single-node` and `--default-bucket` flags for automatic configuration.  If
+you are using an older version of Garage (before v2.3.0), you must follow
+these instructions as automatic configuration is not available.
+
+We will have to run quite a few `garage` administration commands to get started.
+If you ever get lost, don't forget that the `help` command and the `--help` flags can help you anywhere,
 the CLI tool is self-documented! Two examples:

 ```
@@ -180,25 +308,80 @@ garage help
 garage bucket allow --help
 ```

-### Create a bucket
+### Configuring the `garage` CLI
+
+Remember that the `garage` CLI needs to know the path of your `garage.toml` configuration file.
+If it is not in the default location of `/etc/garage.toml`, you can specify it either:
+
+- by setting the `GARAGE_CONFIG_FILE` environment variable;
+- by adding the `-c` flag to each `garage` command, for example: `garage -c ./garage.toml status`.
+
+If you are running Garage in a Docker container, you can set the following alias
+to provide a fake `garage`command that uses the Garage binary inside your container:
+
+```bash
+alias garage="docker exec -ti <container name> /garage"
+```
+
+You can test that your `garage` CLI is configured correctly by running a basic command such as `garage status`.
+
+### Creating a cluster layout
+
+When you first start a cluster without automatic configuration, the output of `garage status` will look as follows:
+
+```
+==== HEALTHY NODES ====
+ID                Hostname  Address         Tags  Zone  Capacity          DataAvail  Version
+563e1ac825ee3323  linuxbox  127.0.0.1:3901              NO ROLE ASSIGNED             v2.3.0
+```
+
+Creating a cluster layout for a Garage deployment means informing Garage of the
+disk space available on each node of the cluster using the `-c` flag, as well
+as the name of the zone (e.g. datacenter) each machine is located in using the
+`-z` flag.
+
+For our test deployment, we are have only one node with zone named `dc1` and a
+capacity of `1G`, though the capacity is ignored for a single node deployment
+and can be changed later when adding new nodes.
+
+```bash
+garage layout assign -z dc1 -c 1G <node_id>
+```
+
+where `<node_id>` corresponds to the identifier of the node shown by `garage status` (first column).
+You can enter simply a prefix of that identifier.
+For instance here you could write just `garage layout assign -z dc1 -c 1G 563e`.
+
+The layout then has to be applied to the cluster, using:
+
+```bash
+garage layout apply --version 1
+```
+
+
+### Creating buckets and keys

 Let's take an example where we want to deploy NextCloud using Garage as the
-main data storage.
+main data storage.  We will suppose that we want to create a bucket named
+`nextcloud-bucket` that will be accessed through a key named
+`nextcloud-app-key`.

-First, create a bucket with the following command:
+#### Create a bucket
+
+First, create the bucket with the following command:

 ```
 garage bucket create nextcloud-bucket
 ```

-Check that everything went well:
+Check that the bucket was created properly:

 ```
 garage bucket list
 garage bucket info nextcloud-bucket
 ```

-### Create an API key
+#### Create an API key

 The `nextcloud-bucket` bucket now exists on the Garage server,
 however it cannot be accessed until we add an API key with the proper access rights.
@@ -209,7 +392,7 @@ one key can access multiple buckets, multiple keys can access one bucket.
 Create an API key using the following command:

 ```
-garage key new --name nextcloud-app-key
+garage key create nextcloud-app-key
 ```

 The output should look as follows:
@@ -221,14 +404,14 @@ Secret key: 7d37d093435a41f2aab8f13c19ba067d9776c90215f56614adad6ece597dbb34
 Authorized buckets:
 ```

-Check that everything works as intended:
+Check that the key was created properly:

 ```
 garage key list
 garage key info nextcloud-app-key
 ```

-### Allow a key to access a bucket
+#### Allow a key to access a bucket

 Now that we have a bucket and a key, we need to give permissions to the key on the bucket:

@@ -247,78 +430,5 @@ You can check at any time the allowed keys on your bucket with:
 garage bucket info nextcloud-bucket
 ```

-
-## Uploading and downlading from Garage
-
-To download and upload files on garage, we can use a third-party tool named `awscli`.
-
-
-### Install and configure `awscli`
-
-If you have python on your system, you can install it with:
-
-```bash
-python -m pip install --user awscli
-```
-
-Now that `awscli` is installed, you must configure it to talk to your Garage instance,
-with your key. There are multiple ways to do that, the simplest one is to create a file
-named `~/.awsrc` with this content:
-
-```bash
-export AWS_ACCESS_KEY_ID=xxxx      # put your Key ID here
-export AWS_SECRET_ACCESS_KEY=xxxx  # put your Secret key here
-export AWS_DEFAULT_REGION='garage'
-export AWS_ENDPOINT_URL='http://localhost:3900'
-
-aws --version
-```
-
-Note you need to have at least `awscli` `>=1.29.0` or `>=2.13.0`, otherwise you
-need to specify `--endpoint-url` explicitly on each `awscli` invocation.
-
-Now, each time you want to use `awscli` on this target, run:
-
-```bash
-source ~/.awsrc
-```
-
-*You can create multiple files with different names if you 
-have multiple Garage clusters or different keys.
-Switching from one cluster to another is as simple as
-sourcing the right file.*
-
-### Example usage of `awscli`
-
-```bash
-# list buckets
-aws s3 ls
-
-# list objects of a bucket
-aws s3 ls s3://nextcloud-bucket
-
-# copy from your filesystem to garage
-aws s3 cp /proc/cpuinfo s3://nextcloud-bucket/cpuinfo.txt
-
-# copy from garage to your filesystem
-aws s3 cp s3://nextcloud-bucket/cpuinfo.txt /tmp/cpuinfo.txt
-```
-
-Note that you can use `awscli` for more advanced operations like
-creating a bucket, pre-signing a request or managing your website.
-[Read the full documentation to know more](https://awscli.amazonaws.com/v2/documentation/api/latest/reference/s3/index.html).
-
-Some features are however not implemented like ACL or policy.
-Check [our s3 compatibility list](@/documentation/reference-manual/s3-compatibility.md).
-
-### Other tools for interacting with Garage
-
-The following tools can also be used to send and recieve files from/to Garage:
-
- [minio-client](@/documentation/connect/cli.md#minio-client) 
- [s3cmd](@/documentation/connect/cli.md#s3cmd) 
- [rclone](@/documentation/connect/cli.md#rclone)
- [Cyberduck](@/documentation/connect/cli.md#cyberduck)
- [WinSCP](@/documentation/connect/cli.md#winscp) 
-
-An exhaustive list is maintained in the ["Integrations" > "Browsing tools" section](@/documentation/connect/_index.md).
+You should now be able to read and write objects to the bucket using the
+credentials created above.
@@ -6,38 +6,167 @@ weight = 40
 The Garage administration API is accessible through a dedicated server whose
 listen address is specified in the `[admin]` section of the configuration
 file (see [configuration file
-reference](@/documentation/reference-manual/configuration.md))
+reference](@/documentation/reference-manual/configuration.md)).

-**WARNING.** At this point, there is no comittement to stability of the APIs described in this document.
-We will bump the version numbers prefixed to each API endpoint at each time the syntax
-or semantics change, meaning that code that relies on these endpoint will break
-when changes are introduced.
+The current version of the admin API is v2.  No breaking changes to the Garage
+administration API will be published outside of a major release.

-The Garage administration API was introduced in version 0.7.2, this document
-does not apply to older versions of Garage.
+History of previous versions:

+ - Before Garage v0.7.2 - no admin API
+ - Garage v0.7.2 - admin API v0
+ - Garage v0.9.0 - admin API v1, deprecate admin API v0
+ - Garage v2.0.0 - admin API v2, deprecate admin API v1

 ## Access control

-The admin API uses two different tokens for acces control, that are specified in the config file's `[admin]` section:
+### Using an API token

- `metrics_token`: the token for accessing the Metrics endpoint (if this token
-  is not set in the config file, the Metrics endpoint can be accessed without
-  access control);
-
- `admin_token`: the token for accessing all of the other administration
-  endpoints (if this token is not set in the config file, access to these
-  endpoints is disabled entirely).
-
-These tokens are used as simple HTTP bearer tokens. In other words, to
-authenticate access to an admin API endpoint, add the following HTTP header
-to your request:
+Administration API tokens tokens are used as simple HTTP bearer tokens. In
+other words, to authenticate access to an admin API endpoint, add the following
+HTTP header to your request:

 ```
 Authorization: Bearer <token>
 ```

-## Administration API endpoints
+### User-defined API tokens
+
+Cluster administrators may dynamically define administration tokens using the CLI commands under `garage admin-token`.
+Such tokens may be limited in scope, meaning that they may enable access to only a subset of API calls.
+They may also have an expiration date to limit their use in time.
+
+Here is an example to create an administration token that is valid for 30 days
+and gives access to only a subset of API calls, allowing it to create buckets
+and access keys and give keys permissions on buckets:
+
+```bash
+$ garage admin-token create --expires-in 30d \
+    --scope ListBuckets,GetBucketInfo,ListKeys,GetKeyInfo,CreateBucket,CreateKey,AllowBucketKey,DenyBucketKey \
+    my-token
+This is your secret bearer token, it will not be shown again by Garage:
+
+  8ed1830b10a276ff57061950.kOSIpxWK9zSGbTO9Xadpv3YndSFWma0_snXcYHaORXk
+
+==== ADMINISTRATION TOKEN INFORMATION ====
+Token ID:    8ed1830b10a276ff57061950
+Token name:  my-token
+Created:     2025-06-15 15:12:44.160 +02:00
+Validity:    valid
+Expiration:  2025-07-15 15:12:44.117 +02:00
+
+Scope:       ListBuckets
+             GetBucketInfo
+             ListKeys
+             GetKeyInfo
+             CreateBucket
+             CreateKey
+             AllowBucketKey
+             DenyBucketKey
+```
+
+When running this command, your token will be shown only once and **will never
+be shown again by Garage**, so make sure to save it directly.  The token is
+hashed internally, and is identified by its prefix (32 hex digits followed by a
+dot) which is saved in clear.
+
+When running `garage admin-token list`, you might see something like this:
+
+```
+ID                        Created     Name                                       Expiration                      Scope
+-                         -           metrics_token (from daemon configuration)  never                           Metrics
+8ed1830b10a276ff57061950  2025-06-15  my-token                                   2025-07-15 15:12:44.117 +02:00  ListBuckets, ... (8)
+```
+
+### Master API tokens
+
+The admin API can also use two different master tokens for access control,
+specified in the config file's `[admin]` section:
+
+- `metrics_token`: the token for accessing the Metrics endpoint. If this token
+  is not set in the config file, the Metrics endpoint can be accessed without
+  access control.
+
+- `admin_token`: the token for accessing all of the other administration
+  endpoints. If this token is not set in the config file, access to these
+  endpoints is only possible with a user-defined admin token.
+
+With the introduction of multiple user-defined admin tokens, the use of master
+API tokens is now discouraged.
+
+
+## Using the admin API
+
+All of the admin API endpoints are described in the OpenAPI specification:
+
+ - APIv2 - [HTML spec](https://garagehq.deuxfleurs.fr/api/garage-admin-v2.html) - [OpenAPI JSON](https://garagehq.deuxfleurs.fr/api/garage-admin-v2.json)
+ - APIv1 (deprecated) - [HTML spec](https://garagehq.deuxfleurs.fr/api/garage-admin-v1.html) - [OpenAPI YAML](https://garagehq.deuxfleurs.fr/api/garage-admin-v1.yml)
+ - APIv0 (deprecated) - [HTML spec](https://garagehq.deuxfleurs.fr/api/garage-admin-v0.html) - [OpenAPI YAML](https://garagehq.deuxfleurs.fr/api/garage-admin-v0.yml)
+
+Making a request to the API from the command line can be as simple as running:
+
+```bash
+curl -H 'Authorization: Bearer s3cr3t' http://localhost:3903/v2/GetClusterStatus | jq
+```
+
+For more advanced use cases, we recommend using an SDK.  
+[Go to the "Build your own app" section to know how to use our SDKs](@/documentation/build/_index.md)
+
+### Making API calls from the `garage` CLI
+
+Since v2.0.0, the `garage` binary provides a subcommand `garage json-api` that
+allows you to invoke the API without making an HTTP request.  This can be
+useful for scripting Garage deployments.
+
+`garage json-api` proxies API calls through Garage's internal RPC protocol,
+therefore it does not require any form of authentication: RPC connection
+parameters are discovered automatically to contact the locally-running Garage
+instance (as when running any other `garage` CLI command).
+
+For simple calls that take no parameters, usage is as follows:
+
+```
+$ garage json-api GetClusterHealth
+{
+  "connectedNodes": 3,
+  "knownNodes": 3,
+  "partitions": 256,
+  "partitionsAllOk": 256,
+  "partitionsQuorum": 256,
+  "status": "healthy",
+  "storageNodes": 3,
+  "storageNodesOk": 3
+}
+```
+
+If you need to specify a JSON body for your call, you can add it directly after
+the name of the function you are calling:
+
+```
+$ garage json-api CreateAdminToken '{"name": "test"}'
+```
+
+Or you can feed it through stdin by adding a `-` as the last command parameter:
+
+```
+$ garage json-api CreateAdminToken -
+{"name": "test"}
+<EOF>
+```
+
+For admin API calls that would have taken query parameters in their HTTP version, these parameters can be passed in the JSON body object:
+
+```
+$ garage json-api GetAdminTokenInfo '{"id":"b0e6e0ace2c0b2aca4cdb2de"}'
+```
+
+For admin API calls that take both query parameters and a JSON body, combine them in the following fashion:
+
+```
+$ garage json-api UpdateAdminToken '{"id":"b0e6e0ace2c0b2aca4cdb2de", "body":{"name":"not a test"}}'
+```
+
+## Special administration API endpoints

 ### Metrics `GET /metrics`

@@ -80,13 +209,13 @@ content-length: 102
 date: Tue, 08 Aug 2023 07:22:38 GMT

 Garage is fully operational
-Consult the full health check API endpoint at /v0/health for more details
+Consult the full health check API endpoint at /v2/GetClusterHealth for more details
 ```

 ### On-demand TLS `GET /check`

-To prevent abuses for on-demand TLS, Caddy developpers have specified an endpoint that can be queried by the reverse proxy
-to know if a given domain is allowed to get a certificate. Garage implements this endpoints to tell if a given domain is handled by Garage or is garbage.
+To prevent abuse for on-demand TLS, Caddy developers have specified an endpoint that can be queried by the reverse proxy
+to know if a given domain is allowed to get a certificate. Garage implements these endpoints to tell if a given domain is handled by Garage or is garbage.

 Garage responds with the following logic:
 - If the domain matches the pattern `<bucket-name>.<s3_api.root_domain>`, returns 200 OK
@@ -99,7 +228,7 @@ You must manually declare the domain in your reverse-proxy. Idem for K2V.*

 *Note 2: buckets in a user's namespace are not supported yet by this endpoint. This is a limitation of this endpoint currently.*

-**Example:** Suppose a Garage instance configured with `s3_api.root_domain = .s3.garage.localhost` and `s3_web.root_domain = .web.garage.localhost`.
+**Example:** Suppose a Garage instance is configured with `s3_api.root_domain = .s3.garage.localhost` and `s3_web.root_domain = .web.garage.localhost`.

 With a private `media` bucket (name in the global namespace, website is disabled), the endpoint will feature the following behavior:

@@ -123,21 +252,7 @@ $ curl -so /dev/null -w "%{http_code}" http://localhost:3903/check?domain=exampl
 200
 ```

-
 **References:**
 - [Using On-Demand TLS](https://caddyserver.com/docs/automatic-https#using-on-demand-tls)
 - [Add option for a backend check to approve use of on-demand TLS](https://github.com/caddyserver/caddy/pull/1939)
 - [Serving tens of thousands of domains over HTTPS with Caddy](https://caddy.community/t/serving-tens-of-thousands-of-domains-over-https-with-caddy/11179)
-
-### Cluster operations
-
-These endpoints are defined on a dedicated [Redocly page](https://garagehq.deuxfleurs.fr/api/garage-admin-v0.html). You can also download its [OpenAPI specification](https://garagehq.deuxfleurs.fr/api/garage-admin-v0.yml).
-
-Requesting the API from the command line can be as simple as running:
-
-```bash
-curl -H 'Authorization: Bearer s3cr3t' http://localhost:3903/v0/status | jq
-```
-
-For more advanced use cases, we recommend using a SDK.  
-[Go to the "Build your own app" section to know how to use our SDKs](@/documentation/build/_index.md)
@@ -37,6 +37,21 @@ A Garage cluster can very easily evolve over time, as storage nodes are added or
 Garage will automatically rebalance data between nodes as needed to ensure the desired number of copies.
 Read about cluster layout management [here](@/documentation/operations/layout.md).

+### Several replication modes
+
+Garage supports a variety of replication modes, with configurable replica count,
+and with various levels of consistency, in order to adapt to a variety of usage scenarios.
+Read our reference page on [supported replication modes](@/documentation/reference-manual/configuration.md#replication_factor)
+to select the replication mode best suited to your use case (hint: in most cases, `replication_factor = 3` is what you want).
+
+### Compression and deduplication
+
+All data stored in Garage is deduplicated, and optionally compressed using
+Zstd.  Objects uploaded to Garage are chunked in blocks of constant sizes (see
+[`block_size`](@/documentation/reference-manual/configuration.md#block_size)),
+and the hashes of individual blocks are used to dispatch them to storage nodes
+and to deduplicate them.
+
 ### No RAFT slowing you down

 It might seem strange to tout the absence of something as a desirable feature,
@@ -46,14 +61,7 @@ directed to a Garage cluster can be handled independently of one another instead
 of going through a central bottleneck (the leader node).
 As a consequence, requests can be handled much faster, even in cases where latency
 between cluster nodes is important (see our [benchmarks](@/documentation/design/benchmarks/index.md) for data on this).
-This is particularly usefull when nodes are far from one another and talk to one other through standard Internet connections.
-
-### Several replication modes
-
-Garage supports a variety of replication modes, with 1 copy, 2 copies or 3 copies of your data,
-and with various levels of consistency, in order to adapt to a variety of usage scenarios.
-Read our reference page on [supported replication modes](@/documentation/reference-manual/configuration.md#replication-mode)
-to select the replication mode best suited to your use case (hint: in most cases, `replication_mode = "3"` is what you want).
+This is particularly useful when nodes are far from one another and talk to one other through standard Internet connections.

 ### Web server for static websites

@@ -76,13 +84,13 @@ exposing the same content under different domain names.

 Garage also supports bucket aliases which are local to a single user:
 this allows different users to have different buckets with the same name, thus avoiding naming collisions.
-This can be helpfull for instance if you want to write an application that creates per-user buckets with always the same name.
+This can be helpful for instance if you want to write an application that creates per-user buckets with always the same name.

 This feature is totally invisible to S3 clients and does not break compatibility with AWS.

 ### Cluster administration API

-Garage provides a fully-fledged REST API to administer your cluster programatically.
+Garage provides a fully-fledged REST API to administer your cluster programmatically.
 Functionality included in the admin API include: setting up and monitoring
 cluster nodes, managing access credentials, and managing storage buckets and bucket aliases.
 A full reference of the administration API is available [here](@/documentation/reference-manual/admin-api.md).
@@ -92,7 +100,7 @@ A full reference of the administration API is available [here](@/documentation/r
 Garage makes some internal metrics available in the Prometheus data format,
 which allows you to build interactive dashboards to visualize the load and internal state of your storage cluster.

-For developpers and performance-savvy administrators,
+For developers and performance-savvy administrators,
 Garage also supports exporting traces of what it does internally in OpenTelemetry format.
 This allows to monitor the time spent at various steps of the processing of requests,
 in order to detect potential performance bottlenecks.
@@ -121,5 +129,5 @@ related to objects stored in an S3 bucket.
 In the context of our research project, [Aérogramme](https://aerogramme.deuxfleurs.fr),
 K2V is used to provide metadata and log storage for operations on encrypted e-mail storage.

-Learn more on the specification of K2V [here](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/k2v/doc/drafts/k2v-spec.md)
+Learn more on the specification of K2V [here](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/commit/f8be15c37db857e177d543de7be863692628d567/doc/drafts/k2v-spec.md)
 and on how to enable it in Garage [here](@/documentation/reference-manual/k2v.md).
@@ -16,10 +16,10 @@ the `k2v` feature flag enabled can be obtained from our download page under
 with `-k2v` (example: `v0.7.2-k2v`).

 The specification of the K2V API can be found
-[here](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/branch/main/doc/drafts/k2v-spec.md).
+[here](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/commit/f8be15c37db857e177d543de7be863692628d567/doc/drafts/k2v-spec.md).
 This document also includes a high-level overview of K2V's design.

-The K2V API uses AWSv4 signatures for authentification, same as the S3 API.
+The K2V API uses AWSv4 signatures for authentication, same as the S3 API.
 The AWS region used for signature calculation is always the same as the one
 defined for the S3 API in the config file.

@@ -55,4 +55,3 @@ cargo build --features cli --bin k2v-cli
 The CLI utility is self-documented, run `k2v-cli --help` to learn how to use
 it. There is also a short README.md in the `src/k2v-client` folder with some
 instructions.
-
@@ -0,0 +1,188 @@
+++
+title = "Known issues"
+weight = 80
+++
+
+Issues in each section are roughly sorted by order of decreasing impact, based on actual reports from  users.
+
+## Architectural limitations
+
+Issues that are caused by design decisions of Garage internals, and that can't
+be fixed without major architectural changes in the codebase.
+
+### Metadata performance issues with many objects
+
+**Related issues:**
+
+- [#851 - Performances collapse with 10 millions pictures in a bucket](https://git.deuxfleurs.fr/Deuxfleurs/garage/issues/851)
+- [#1222 - Cluster Setup Write Performance Degraded After Writing 10 Million Object (200-300Kb per object)](https://git.deuxfleurs.fr/Deuxfleurs/garage/issues/1222)
+
+### Very big objects cause performance degradation
+
+For each object, there is a single metadata entry called a `Version` that
+contains a list of all of the data blocks in the object.  For very big objects,
+this entry can contain thousands of block references.  During the uploading of
+an object, this metadata entry needs to be read, deserialized, reserialized and
+written for each individual data block uploaded.  This means that the
+complexity of an upload is `O(n²)` in the number of blocks needed.
+
+This manifests by excessive metadata I/O and CPU usage, and uploads eventually stalling.
+
+**Mitigation:** Increase the `block_size` configuration parameter to reduce the
+number of blocks. Make sure multipart uploads use chunks that are at least
+`block_size` in size, and that are an exact multiple of `block_size` to avoid
+the creation of smaller blocks.
+
+**Long-term solution:** An architectural change in the metadata system would be
+required to store block lists in many independent metadata entries instead of
+one single big entry per object.
+
+**Related issues:**
+
+- [#662 - Large Files fail to upload](https://git.deuxfleurs.fr/Deuxfleurs/garage/issues/662)
+- [#1366 - High CPU usage and performance degradation during long multipart uploads](https://git.deuxfleurs.fr/Deuxfleurs/garage/issues/1366)
+
+### No conditional writes / locking / WORM support (`if-none-match`, ...)
+
+This is structurally impossible to implement in Garage due to the lack of a consensus algorithm,
+which is one of Garage's core design choices which we cannot reconsider.
+
+A semi-working, *unsafe* implementation of WORM and object locking could be
+implemented, with the following constraint: only after the completion of the
+first write (in case of WORM) or the setting of a lock (for object lock) can we
+guarantee that the object cannot be overwritten. In case where an overwrite
+requests arrives at the same time as the initial request to write or to lock
+the object, we cannot implement a safe and consistent way to reject it.  This
+means that many practical use-cases for `if-none-match` cannot be supported
+(e.g. using it to implement mutual exclusion between concurrent writers).
+
+**Related issues:**
+
+- [#1052 - Support conditional writes](https://git.deuxfleurs.fr/Deuxfleurs/garage/issues/1052)
+- [#1127 - Feature Request: WORM (Write Once Read Many) / Object Lock Support](https://git.deuxfleurs.fr/Deuxfleurs/garage/issues/1127)
+
+### `CreateBucket` race condition
+
+Also due to the lack of a consensus algorithm, there is no mutual exclusion
+between concurrent `CreateBucket` requests using the same bucket name.
+
+**Related issues:**
+
+- [#649 - Race condition in CreateBucket](https://git.deuxfleurs.fr/Deuxfleurs/garage/issues/649)
+
+### Metadata and data have the same replication factor
+
+There is a single `replication_factor` in the configuration file that applies both to data blocks and metadata entries.
+This makes clusters with `replication_factor = 1` particularly vulnerable in cases of metadata corruption (see below), as there
+is a single copy of the metadata for each object even in multi-node clusters.
+
+**Mitigation:** Do not use `replication_factor = 1`.
+
+**Long-term solution:** We want to allow scenarios such as replicating the
+metadata on 2, 3 or more nodes and the data on only 1 or 2 nodes (for example),
+so that the metadata can benefit from better redundancy without increasing the
+storage costs for the entire dataset. This will require some important changes
+in the codebase.
+
+**Related issues:**
+
+- [#720 - Separate replication modes for metadata/data](https://git.deuxfleurs.fr/Deuxfleurs/garage/issues/720)
+
+### Node count limitation
+
+Garage will have issues in clusters with too many nodes, it will not be able to
+spread data uniformly among nodes and some nodes will fill up faster than
+other. This starts to manifest when the number of nodes is bigger than `10 ×
+replication_factor`.  This is due to the fact that Garage uses only 256
+partitions internally.
+
+**Mitigation:** Build clusters with fewer, bigger nodes.
+
+**Potential solution:** This can be fixed by increasing the number of
+partitions in Garage. The code paths exist, there is [a `const`
+somewhere](https://git.deuxfleurs.fr/Deuxfleurs/garage/src/commit/6fd9bba0cb55062cb1725ab961b7fa8acb9dcc61/src/rpc/layout/mod.rs#L35)
+that theoretically allows to increase the number of partitions up to `2^16`,
+but this has not been tested so there might be bugs.
+
+### Buckets are not sharded
+
+For each bucket, the first metadata layer that contains an index of all objects
+is not sharded.  This index, which includes the names and all metadata (size,
+headers, ...) for each object, is stored on `$replication_factor` nodes.
+
+For instance with `replication_factor = 3`, a given bucket will use only 3
+specific nodes for this index (chosen at random when the bucket is created) to
+store this index.  In a multi-zone deployments, these nodes will be spread in
+different zones.  Each bucket uses a different set of 3 random nodes for its
+index.
+
+As a consequence, very large buckets might cause uneven load distribution
+within a cluster.  If all of the requests on a cluster are for objects in a
+single bucket, then the `$replication_factor` nodes that store the index will
+become a hotspot in the cluster, with more intensive metadata access patterns.
+There is no way of choosing which nodes will have this role.
+
+Currently, we have no report of this being an issue in practice.
+
+**Mitigation:** This impacts in particular clusters that are used for a single
+purpose with a single bucket. This can be solved by dividing your dataset among
+many buckets, using a client-side sharding strategy that you will have to
+design. Use at least as many buckets as you have nodes on your cluster.
+
+
+## Bugs
+
+Known bugs that are complex to diagnose and fix, and therefore have not been
+fixed yet.
+
+### LMDB metadata corruption
+
+Many users have reported situations where the LMDB metadata db becomes
+corrupted, sometimes after a forced shutdown of Garage or in case of power
+loss.  A corrupted database file is generally not recoverable.
+
+**Mitigation:** Use a `replication_factor` of at least 2. Configure automatic
+snapshotting using `metadata_auto_snapshot_interval` so that in case of
+corruption you can rollback to a working database.
+
+Note that taking filesystem-level snapshots of your `metadata_dir`, although it
+is much faster and less I/O intensive than Garage's built-in snapshotting, does
+not ensure that the snapshot will be consistent. If the snapshot is taking
+during a metadata write, the snapshot itself might be corrupted and thus not
+usable as a rollback point. Therefore, prefer using
+`metadata_auto_snapshot_interval` in all cases.
+
+### Layout updates might require manual intervention
+
+In case of disconnected nodes, when changing the cluster layout to remove these
+nodes and add other nodes instead, Garage might not be able to properly evict
+the old nodes from the system. This is a built-in security measure to avoid any
+inconsistent cluster states.
+
+This manifests by several cluster layout versions staying active even after a
+full resync. You can diagnose this situation with `garage layout history`,
+which will give you instructions to fix it.
+
+### Tag assignment
+
+In the `garage layout assign` command, the `-t` argument has to be repeated
+multiple times to set multiple tags on a node. Writing multiple tags separated
+by commas will result in a single string.
+
+## General footguns
+
+Choices made by the developers that users must be aware of if they don't want
+to run into potential issues.
+
+### Resync tranquility is conservative by default
+
+By default, the worker parameters `resync-tranquility` and `resync-worker-count` are set to very conservative values, to avoid overloading nodes with I/O when data needs to be resynchronized between nodes.
+This can cause issues where the resync queue grows faster than it can be cleared, which in turn causes performance issues in the rest of Garage.
+
+This situation is indicated by a big resync queue with few resync errors (the queue is not caused by a disconnected/malfunctionning node).
+To fix it, increase the number of resync workers and reduce the resync tranquility. For instance, if you want to resync as fast as possible:
+
+```
+garage worker set -a resync-worker-count 8
+garage worker set -a resync-tranquility 0
+```
@@ -27,6 +27,112 @@ Exposes the Garage replication factor configured on the node
 garage_replication_factor 3
 ```

+#### `garage_local_disk_avail` and `garage_local_disk_total` (gauge)
+
+Reports the available and total disk space on each node, for data and metadata separately.
+
+```
+garage_local_disk_avail{volume="data"} 540341960704
+garage_local_disk_avail{volume="metadata"} 540341960704
+garage_local_disk_total{volume="data"} 763063566336
+garage_local_disk_total{volume="metadata"} 763063566336
+```
+
+### Cluster health status metrics
+
+#### `cluster_healthy` (gauge)
+
+Whether all storage nodes are connected (0 or 1)
+
+```
+cluster_healthy 0
+```
+
+#### `cluster_available` (gauge)
+
+Whether all requests can be served, even if some storage nodes are disconnected
+
+```
+cluster_available 1
+```
+
+#### `cluster_connected_nodes` (gauge)
+
+Number of nodes currently connected
+
+```
+cluster_connected_nodes 3
+```
+
+#### `cluster_known_nodes` (gauge)
+
+Number of nodes already seen once in the cluster
+
+```
+cluster_known_nodes 3
+```
+
+#### `cluster_layout_node_connected` (gauge)
+
+Connection status for individual nodes of the cluster layout
+
+```
+cluster_layout_node_connected{id="62b218d848e86a64",role_capacity="1000000000",role_gateway="0",role_zone="dc1"} 1
+cluster_layout_node_connected{id="a11c7cf18af29737",role_capacity="1000000000",role_gateway="0",role_zone="dc1"} 0
+cluster_layout_node_connected{id="a235ac7695e0c54d",role_capacity="1000000000",role_gateway="0",role_zone="dc1"} 1
+cluster_layout_node_connected{id="b10c110e4e854e5a",role_capacity="1000000000",role_gateway="0",role_zone="dc1"} 1
+```
+
+#### `cluster_layout_node_disconnected_time` (gauge)
+
+Time (in seconds) since last connection to individual nodes of the cluster layout
+
+```
+cluster_layout_node_disconnected_time{id="62b218d848e86a64",role_capacity="1000000000",role_gateway="0",role_zone="dc1"} 0
+cluster_layout_node_disconnected_time{id="a235ac7695e0c54d",role_capacity="1000000000",role_gateway="0",role_zone="dc1"} 0
+cluster_layout_node_disconnected_time{id="b10c110e4e854e5a",role_capacity="1000000000",role_gateway="0",role_zone="dc1"} 0
+```
+
+#### `cluster_storage_nodes` (gauge)
+
+Number of storage nodes declared in the current layout
+
+```
+cluster_storage_nodes 4
+```
+
+#### `cluster_storage_nodes_ok` (gauge)
+
+Number of storage nodes currently connected
+
+```
+cluster_storage_nodes_ok 3
+```
+
+#### `cluster_partitions` (gauge)
+
+Number of partitions in the layout (this is always 256)
+
+```
+cluster_partitions 256
+```
+
+#### `cluster_partitions_all_ok` (gauge)
+
+Number of partitions for which all storage nodes are connected
+
+```
+cluster_partitions_all_ok 64
+```
+
+#### `cluster_partitions_quorum` (gauge)
+
+Number of partitions for which we have a quorum of connected nodes and all requests can be served
+
+```
+cluster_partitions_quorum 256
+```
+
 ### Metrics of the API endpoints

 #### `api_admin_request_counter` (counter)
@@ -119,6 +225,17 @@ block_bytes_read 120586322022
 block_bytes_written 3386618077
 ```

+#### `block_ram_buffer_free_kb` (gauge)
+
+Kibibytes available for buffering blocks that have to be sent to remote nodes.
+When clients send too much data to this node and a storage node is not receiving
+data fast enough due to slower network conditions, this will decrease down to
+zero and backpressure will be applied.
+
+```
+block_ram_buffer_free_kb 219829
+```
+
 #### `block_compression_level` (counter)

 Exposes the block compression level configured for the Garage node.
@@ -275,7 +392,7 @@ table_merkle_updater_todo_queue_length{table_name="block_ref"} 0

 #### `table_sync_items_received`, `table_sync_items_sent` (counters)

-Number of data items sent to/recieved from other nodes during resync procedures
+Number of data items sent to/received from other nodes during resync procedures

 ```
 table_sync_items_received{from="<remote node>",table_name="bucket_v2"} 3
@@ -23,16 +23,17 @@ Feel free to open a PR to suggest fixes this table. Minio is missing because the
 - 2022-05-25 - Many Ceph S3 endpoints are not documented but implemented. Following a notification from the Ceph community, we added them.


-
 ## High-level features

 | Feature                      | Garage                           | [Openstack Swift](https://docs.openstack.org/swift/latest/s3_compat.html) | [Ceph Object Gateway](https://docs.ceph.com/en/latest/radosgw/s3/) | [Riak CS](https://docs.riak.com/riak/cs/2.1.1/references/apis/storage/s3/index.html) | [OpenIO](https://docs.openio.io/latest/source/arch-design/s3_compliancy.html) |
 |------------------------------|----------------------------------|-----------------|---------------|---------|-----|
-| [signature v2](https://docs.aws.amazon.com/general/latest/gr/signature-version-2.html) (deprecated) | ❌ Missing | ✅ |  ✅ | ✅ | ✅ |
+| [signature v2](https://docs.aws.amazon.com/AmazonS3/latest/API/Appendix-Sigv2.html) (deprecated) | ❌ Missing | ✅ |  ✅ | ✅ | ✅ |
 | [signature v4](https://docs.aws.amazon.com/AmazonS3/latest/API/sig-v4-authenticating-requests.html) |  ✅ Implemented |  ✅ |  ✅ | ❌ | ✅ |
 | [URL path-style](https://docs.aws.amazon.com/AmazonS3/latest/userguide/VirtualHosting.html#path-style-access) (eg. `host.tld/bucket/key`) |  ✅ Implemented | ✅ |  ✅ | ❓| ✅ |
 | [URL vhost-style](https://docs.aws.amazon.com/AmazonS3/latest/userguide/VirtualHosting.html#virtual-hosted-style-access) URL (eg. `bucket.host.tld/key`) |  ✅ Implemented | ❌| ✅| ✅ | ✅ |
 | [Presigned URLs](https://docs.aws.amazon.com/AmazonS3/latest/userguide/ShareObjectPreSignedURL.html) |  ✅ Implemented | ❌|  ✅ | ✅ |  ✅(❓) |
+| [SSE-C encryption](https://docs.aws.amazon.com/AmazonS3/latest/userguide/ServerSideEncryptionCustomerKeys.html) |  ✅ Implemented | ❓ |  ✅ | ❌ |  ✅ |
+| [Bucket versioning](https://docs.aws.amazon.com/AmazonS3/latest/userguide/Versioning.html) | ❌ Missing | ✅ |  ✅ | ❌ | ✅ |

 *Note:* OpenIO does not says if it supports presigned URLs. Because it is part
 of signature v4 and they claim they support it without additional precisions,
@@ -44,7 +45,7 @@ we suppose that OpenIO supports presigned URLs.
 All endpoints that are missing on Garage will return a 501 Not Implemented.
 Some `x-amz-` headers are not implemented.

-### Core endoints
+### Core endpoints

 | Endpoint                     | Garage                           | [Openstack Swift](https://docs.openstack.org/swift/latest/s3_compat.html) | [Ceph Object Gateway](https://docs.ceph.com/en/latest/radosgw/s3/) | [Riak CS](https://docs.riak.com/riak/cs/2.1.1/references/apis/storage/s3/index.html) | [OpenIO](https://docs.openio.io/latest/source/arch-design/s3_compliancy.html) |
 |------------------------------|----------------------------------|-----------------|---------------|---------|-----|
@@ -75,16 +76,13 @@ but these endpoints are documented in [Red Hat Ceph Storage - Chapter 2. Ceph Ob

 | Endpoint                     | Garage                           | [Openstack Swift](https://docs.openstack.org/swift/latest/s3_compat.html) | [Ceph Object Gateway](https://docs.ceph.com/en/latest/radosgw/s3/) | [Riak CS](https://docs.riak.com/riak/cs/2.1.1/references/apis/storage/s3/index.html) | [OpenIO](https://docs.openio.io/latest/source/arch-design/s3_compliancy.html) |
 |------------------------------|----------------------------------|-----------------|---------------|---------|-----|
-| [AbortMultipartUpload](https://docs.aws.amazon.com/AmazonS3/latest/API/API_AbortMultipartUpload.html)         | ✅ Implemented                      |  ✅ | ✅ | ✅ | ✅ |
-| [CompleteMultipartUpload](https://docs.aws.amazon.com/AmazonS3/latest/API/API_CompleteMultipartUpload.html)   | ✅ Implemented (see details below)                      | ✅ | ✅ | ✅ | ✅ |
-| [CreateMultipartUpload](https://docs.aws.amazon.com/AmazonS3/latest/API/API_CreateMultipartUpload.html)        | ✅ Implemented                      | ✅| ✅ | ✅ | ✅ |
-| [ListMultipartUpload](https://docs.aws.amazon.com/AmazonS3/latest/API/API_ListMultipartUpload.html)          | ✅ Implemented                      | ✅ | ✅ | ✅ | ✅ |
-| [ListParts](https://docs.aws.amazon.com/AmazonS3/latest/API/API_ListParts.html)                    | ✅ Implemented                      | ✅ | ✅ | ✅ | ✅ |
-| [UploadPart](https://docs.aws.amazon.com/AmazonS3/latest/API/API_UploadPart.html)                  | ✅ Implemented (see details below)                      | ✅ | ✅| ✅ | ✅ |
-| [UploadPartCopy](https://docs.aws.amazon.com/AmazonS3/latest/API/API_UploadPartCopy.html)               | ✅ Implemented                      | ✅ | ✅ | ✅ | ✅ |
-
-Our implementation of Multipart Upload is currently a bit more restrictive than Amazon's one in some edge cases.
-For more information, please refer to our [issue tracker](https://git.deuxfleurs.fr/Deuxfleurs/garage/issues/204).
+| [AbortMultipartUpload](https://docs.aws.amazon.com/AmazonS3/latest/API/API_AbortMultipartUpload.html)         | ✅ Implemented  |  ✅ | ✅ | ✅ | ✅ |
+| [CompleteMultipartUpload](https://docs.aws.amazon.com/AmazonS3/latest/API/API_CompleteMultipartUpload.html)   | ✅ Implemented  | ✅ | ✅ | ✅ | ✅ |
+| [CreateMultipartUpload](https://docs.aws.amazon.com/AmazonS3/latest/API/API_CreateMultipartUpload.html)        | ✅ Implemented | ✅| ✅ | ✅ | ✅ |
+| [ListMultipartUpload](https://docs.aws.amazon.com/AmazonS3/latest/API/API_ListMultipartUpload.html)          | ✅ Implemented   | ✅ | ✅ | ✅ | ✅ |
+| [ListParts](https://docs.aws.amazon.com/AmazonS3/latest/API/API_ListParts.html)                    | ✅ Implemented             | ✅ | ✅ | ✅ | ✅ |
+| [UploadPart](https://docs.aws.amazon.com/AmazonS3/latest/API/API_UploadPart.html)                  | ✅ Implemented             | ✅ | ✅| ✅ | ✅ |
+| [UploadPartCopy](https://docs.aws.amazon.com/AmazonS3/latest/API/API_UploadPartCopy.html)               | ✅ Implemented        | ✅ | ✅ | ✅ | ✅ |

 ### Website endpoints

@@ -127,15 +125,22 @@ If you need this feature, please [share your use case in our dedicated issue](ht

 | Endpoint                     | Garage                           | [Openstack Swift](https://docs.openstack.org/swift/latest/s3_compat.html) | [Ceph Object Gateway](https://docs.ceph.com/en/latest/radosgw/s3/) | [Riak CS](https://docs.riak.com/riak/cs/2.1.1/references/apis/storage/s3/index.html) | [OpenIO](https://docs.openio.io/latest/source/arch-design/s3_compliancy.html) |
 |------------------------------|----------------------------------|-----------------|---------------|---------|-----|
-| [DeleteBucketLifecycle](https://docs.aws.amazon.com/AmazonS3/latest/API/API_DeleteBucketLifecycle.html) | ❌ Missing | ❌| ✅| ❌| ✅|
-| [GetBucketLifecycleConfiguration](https://docs.aws.amazon.com/AmazonS3/latest/API/API_GetBucketLifecycleConfiguration.html) | ❌ Missing | ❌| ✅ | ❌| ✅|
-| [PutBucketLifecycleConfiguration](https://docs.aws.amazon.com/AmazonS3/latest/API/API_PutBucketLifecycleConfiguration.html) | ❌ Missing | ❌| ✅ | ❌| ✅|
+| [DeleteBucketLifecycle](https://docs.aws.amazon.com/AmazonS3/latest/API/API_DeleteBucketLifecycle.html) | ✅ Implemented | ❌| ✅| ❌| ✅|
+| [GetBucketLifecycleConfiguration](https://docs.aws.amazon.com/AmazonS3/latest/API/API_GetBucketLifecycleConfiguration.html) | ✅ Implemented | ❌| ✅ | ❌| ✅|
+| [PutBucketLifecycleConfiguration](https://docs.aws.amazon.com/AmazonS3/latest/API/API_PutBucketLifecycleConfiguration.html) | ⚠ Partially implemented (see below) | ❌| ✅ | ❌| ✅|
 | [GetBucketVersioning](https://docs.aws.amazon.com/AmazonS3/latest/API/API_GetBucketVersioning.html)          | ❌ Stub (see below)       | ✅| ✅ | ❌| ✅|
 | [ListObjectVersions](https://docs.aws.amazon.com/AmazonS3/latest/API/API_ListObjectVersions.html) | ❌ Missing | ❌| ✅ | ❌| ✅|
 | [PutBucketVersioning](https://docs.aws.amazon.com/AmazonS3/latest/API/API_PutBucketVersioning.html) | ❌ Missing | ❌| ✅| ❌| ✅|

+**PutBucketLifecycleConfiguration:** The only actions supported are
+`AbortIncompleteMultipartUpload` and `Expiration` (without the
+`ExpiredObjectDeleteMarker` field).  All other operations are dependent on
+either bucket versioning or storage classes which Garage currently does not
+implement. The deprecated `Prefix` member directly in the the `Rule`
+structure/XML tag is not supported, specified prefixes must be inside the
+`Filter` structure/XML tag.

-**GetBucketVersioning:** Stub implementation (Garage does not yet support versionning so this always returns "versionning not enabled").
+**GetBucketVersioning:** Stub implementation which always returns "versioning not enabled", since Garage does not yet support bucket versioning.

 ### Replication endpoints

@@ -150,7 +155,7 @@ Please open an issue if you have a use case for replication.
 *Note: Ceph documentation briefly says that Ceph supports
 [replication through the S3 API](https://docs.ceph.com/en/latest/radosgw/multisite-sync-policy/#s3-replication-api)
 but with some limitations.
-Additionaly, replication endpoints are not documented in the S3 compatibility page so I don't know what kind of support we can expect.*
+Additionally, replication endpoints are not documented in the S3 compatibility page so I don't know what kind of support we can expect.*

 ### Locking objects

@@ -192,7 +197,7 @@ Please open an issue if you have a use case.

 ### Vendor specific endpoints

-<details><summary>Display Amazon specifc endpoints</summary>
+<details><summary>Display Amazon specific endpoints</summary>


 | Endpoint                     | Garage                           | [Openstack Swift](https://docs.openstack.org/swift/latest/s3_compat.html) | [Ceph Object Gateway](https://docs.ceph.com/en/latest/radosgw/s3/) | [Riak CS](https://docs.riak.com/riak/cs/2.1.1/references/apis/storage/s3/index.html) | [OpenIO](https://docs.openio.io/latest/source/arch-design/s3_compliancy.html) |
@@ -229,4 +234,3 @@ Please open an issue if you have a use case.
 | [SelectObjectContent](https://docs.aws.amazon.com/AmazonS3/latest/API/API_SelectObjectContent.html) | ❌ Missing | ❌| ❌| ❌| ❌|

 </details>
-
@@ -3,7 +3,7 @@ title = "S3 compatibility target"
 weight = 5
 +++

-If there is a specific S3 functionnality you have a need for, feel free to open
+If there is a specific S3 functionality you have a need for, feel free to open
 a PR to put the corresponding endpoints higher in the list.  Please explain
 your motivations for doing so in the PR message.

@@ -42,7 +42,7 @@ The general principle are similar, but details have not been updated.**
 A version is defined by the existence of at least one entry in the blocks table for a certain version UUID.
 We must keep the following invariant: if a version exists in the blocks table, it has to be referenced in the objects table.
 We explicitly manage concurrent versions of an object: the version timestamp and version UUID columns are index columns, thus we may have several concurrent versions of an object.
-Important: before deleting an older version from the objects table, we must make sure that we did a successfull delete of the blocks of that version from the blocks table.
+Important: before deleting an older version from the objects table, we must make sure that we did a successful delete of the blocks of that version from the blocks table.

 Thus, the workflow for reading an object is as follows:

@@ -68,7 +68,7 @@ Workflow for DELETE:
 1. Check write permission (LDAP)
 2. Get current version (or versions) in object table
 3. Do the deletion of those versions NOT IN A BACKGROUND JOB THIS TIME
-4. Return succes to the user if we were able to delete blocks from the blocks table and entries from the object table
+4. Return success to the user if we were able to delete blocks from the blocks table and entries from the object table

 To delete a version:

@@ -92,10 +92,10 @@ Known issue: if someone is reading from a version that we want to delete and the
 - file path = /meta/(first 3 hex digits of hash)/(rest of hash)
 - map block hash -> set of version UUIDs where it is referenced

-Usefull metadata:
+Useful metadata:

 - list of versions that reference this block in the Casandra table, so that we can do GC by checking in Cassandra that the lines still exist
- list of other nodes that we know have acknowledged a write of this block, usefull in the rebalancing algorithm
+- list of other nodes that we know have acknowledged a write of this block, useful in the rebalancing algorithm

 Write strategy: have a single thread that does all write IO so that it is serialized (or have several threads that manage independent parts of the hash space). When writing a blob, write it to a temporary file, close, then rename so that a concurrent read gets a consistent result (either not found or found with whole content).

@@ -49,12 +49,12 @@ The ring construction that selects `n_token` random positions for each nodes giv
 is not well-balanced: the space between the tokens varies a lot, and some partitions are thus bigger than others.
 This problem was demonstrated in the original Dynamo DB paper.

-To solve this, we want to apply a better second method for partitionning our dataset:
+To solve this, we want to apply a better second method for partitioning our dataset:

 1. fix an initially large number of partitions (say 1024) with evenly-spaced delimiters, 

 2. attribute each partition randomly to a node, with a probability
-   proportionnal to its capacity (which `n_tokens` represented in the first
+   proportional to its capacity (which `n_tokens` represented in the first
   method)

 For now we continue using the multi-DC ring walking described above.
@@ -66,7 +66,7 @@ I have studied two ways to do the attribution of partitions to nodes, in a way t

 MagLev provided significantly better balancing, as it guarantees that the exact
 same number of partitions is attributed to all nodes that have the same
-capacity (and that this number is proportionnal to the node's capacity, except
+capacity (and that this number is proportional to the node's capacity, except
 for large values), however in both cases:

 - the distribution is still bad, because we use the naive multi-DC ring walking
@@ -1,6 +1,6 @@
 +++
 title = "Migrating from 0.3 to 0.4"
-weight = 20
+weight = 80
 +++

 **Migrating from 0.3 to 0.4 is unsupported. This document is only intended to
@@ -68,7 +68,7 @@ The migration steps are as follows:
 5. Turn off Garage 0.3

 6. Backup metadata folders if you can (i.e. if you have space to do it
-   somewhere). Backuping data folders could also be usefull but that's much
+   somewhere). Backuping data folders could also be useful but that's much
   harder to do. If your filesystem supports snapshots, this could be a good
   time to use them.

@@ -1,6 +1,6 @@
 +++
 title = "Migrating from 0.5 to 0.6"
-weight = 15
+weight = 75
 +++

 **This guide explains how to migrate to 0.6 if you have an existing 0.5 cluster.
@@ -1,6 +1,6 @@
 +++
 title = "Migrating from 0.6 to 0.7"
-weight = 14
+weight = 74
 +++
 **This guide explains how to migrate to 0.7 if you have an existing 0.6 cluster.
 We don't recommend trying to migrate to 0.7 directly from 0.5 or older.**
@@ -19,7 +19,7 @@ The migration steps are as follows:
 2. Disable API and web access. Garage does not support disabling
   these endpoints but you can change the port number or stop your reverse
   proxy for instance.
-3. Check once again that your cluster is healty. Run again `garage repair --all-nodes --yes tables` which is quick.
+3. Check once again that your cluster is healthy. Run again `garage repair --all-nodes --yes tables` which is quick.
   Also check your queues are empty, run `garage stats` to query them.
 4. Turn off Garage v0.6
 5. Backup the metadata folder of all your nodes: `cd /var/lib/garage ; tar -acf meta-v0.6.tar.zst meta/`
@@ -1,6 +1,6 @@
 +++
 title = "Migrating from 0.7 to 0.8"
-weight = 13
+weight = 73
 +++

 **This guide explains how to migrate to 0.8 if you have an existing 0.7 cluster.
@@ -0,0 +1,72 @@
+++
+title = "Migrating from 0.8 to 0.9"
+weight = 72
+++
+
+**This guide explains how to migrate to 0.9 if you have an existing 0.8 cluster.
+We don't recommend trying to migrate to 0.9 directly from 0.7 or older.**
+
+This migration procedure has been tested on several clusters without issues.
+However, it is still a *critical procedure* that might cause issues.
+**Make sure to back up all your data before attempting it!**
+
+You might also want to read our [general documentation on upgrading Garage](@/documentation/operations/upgrading.md).
+
+The following are **breaking changes** in Garage v0.9 that require your attention when migrating:
+
+- LMDB is now the default metadata db engine and Sled is deprecated. If you were using Sled, make sure to specify `db_engine = "sled"` in your configuration file, or take the time to [convert your database](https://garagehq.deuxfleurs.fr/documentation/reference-manual/configuration/#db-engine-since-v0-8-0).
+
+- Capacity values are now in actual byte units. The translation from the old layout will assign 1 capacity = 1Gb by default, which might be wrong for your cluster. This does not cause any data to be moved around, but you might want to re-assign correct capacity values post-migration.
+
+- Multipart uploads that were started in Garage v0.8 will not be visible in Garage v0.9 and will have to be restarted from scratch.
+
+- Changes to the admin API: some `v0/` endpoints have been replaced by `v1/` counterparts with updated/uniformized syntax. All other endpoints have also moved to `v1/` by default, without syntax changes, but are still available under `v0/` for compatibility.
+
+
+## Simple migration procedure (takes cluster offline for a while)
+
+The migration steps are as follows:
+
+1. Disable API and web access. You may do this by stopping your reverse proxy or by commenting out
+   the `api_bind_addr` values in your `config.toml` file and restarting Garage.
+2. Do `garage repair --all-nodes --yes tables` and `garage repair --all-nodes --yes blocks`,
+   check the logs and check that all data seems to be synced correctly between
+   nodes. If you have time, do additional checks (`versions`, `block_refs`, etc.)
+3. Check that the block resync queue and Merkle queue are empty:
+   run `garage stats -a` to query them or inspect metrics in the Grafana dashboard.
+4. Turn off Garage v0.8
+5. **Backup the metadata folder of all your nodes!** For instance, use the following command
+	if your metadata directory is `/var/lib/garage/meta`: `cd /var/lib/garage ; tar -acf meta-v0.8.tar.zst meta/`
+6. Install Garage v0.9
+7. Update your configuration file if necessary.
+8. Turn on Garage v0.9
+9. Do `garage repair --all-nodes --yes tables` and `garage repair --all-nodes --yes blocks`.
+   Wait for a full table sync to run.
+10. Your upgraded cluster should be in a working state. Re-enable API and Web
+    access and check that everything went well.
+11. Monitor your cluster in the next hours to see if it works well under your production load, report any issue.
+12. You might want to assign correct capacity values to all your nodes. Doing so might cause data to be moved
+    in your cluster, which should also be monitored carefully.
+
+## Minimal downtime migration procedure
+
+The migration to Garage v0.9 can be done with almost no downtime,
+by restarting all nodes at once in the new version.
+
+The migration steps are as follows:
+
+1. Do `garage repair --all-nodes --yes tables` and `garage repair --all-nodes --yes blocks`,
+   check the logs and check that all data seems to be synced correctly between
+   nodes. If you have time, do additional checks (`versions`, `block_refs`, etc.)
+
+2. Turn off each node individually; back up its metadata folder (see above); turn it back on again.
+   This will allow you to take a backup of all nodes without impacting global cluster availability.
+   You can do all nodes of a single zone at once as this does not impact the availability of Garage.
+
+3. Prepare your binaries and configuration files for Garage v0.9
+
+4. Shut down all v0.8 nodes simultaneously, and restart them all simultaneously in v0.9.
+   Use your favorite deployment tool (Ansible, Kubernetes, Nomad) to achieve this as fast as possible.
+   Garage v0.9 should be in a working state as soon as it starts.
+
+5. Proceed with repair and monitoring as described in steps 9-12 above.
@@ -0,0 +1,77 @@
+++
+title = "Migrating from 0.9 to 1.0"
+weight = 71
+++
+
+**This guide explains how to migrate to 1.0 if you have an existing 0.9 cluster.
+We don't recommend trying to migrate to 1.0 directly from 0.8 or older.**
+
+This migration procedure has been tested on several clusters without issues.
+However, it is still a *critical procedure* that might cause issues.
+**Make sure to back up all your data before attempting it!**
+
+You might also want to read our [general documentation on upgrading Garage](@/documentation/operations/upgrading.md).
+
+## Changes introduced in v1.0
+
+The following are **breaking changes** in Garage v1.0 that require your attention when migrating:
+
+- The Sled metadata db engine has been **removed**. If your cluster was still
+  using Sled, you will need to **use a Garage v0.9.x binary** to convert the
+  database using the `garage convert-db` subcommand. See
+  [here](@/documentation/reference-manual/configuration.md#db_engine) for the
+  details of the procedure.
+
+The following syntax changes have been made to the configuration file:
+
+- The `replication_mode` parameter has been split into two parameters:
+  [`replication_factor`](@/documentation/reference-manual/configuration.md#replication_factor)
+  and
+  [`consistency_mode`](@/documentation/reference-manual/configuration.md#consistency_mode).
+  The old syntax using `replication_mode` is still supported for legacy
+  reasons and can still be used.
+
+- The parameters `sled_cache_capacity` and `sled_flush_every_ms` have been removed.
+
+## Migration procedure
+
+The migration to Garage v1.0 can be done with almost no downtime,
+by restarting all nodes at once in the new version.
+
+The migration steps are as follows:
+
+1. Do a `garage repair --all-nodes --yes tables`, check the logs and check that
+   all data seems to be synced correctly between nodes. If you have time, do
+   additional `garage repair` procedures (`blocks`, `versions`, `block_refs`,
+   etc.)
+
+2. Ensure you have a snapshot of your Garage installation that you can restore
+   to in case the upgrade goes wrong:
+
+   - If you are running Garage v0.9.4 or later, use the `garage meta snapshot
+     --all` to make a backup snapshot of the metadata directories of your nodes
+     for backup purposes, and save a copy of the following files in the
+     metadata directories of your nodes: `cluster_layout`, `data_layout`,
+     `node_key`, `node_key.pub`.
+
+   - If you are running a filesystem such as ZFS or BTRFS that support
+     snapshotting, you can create a filesystem-level snapshot to be used as a
+     restoration point if needed.
+
+   - In other cases, make a backup using the old procedure: turn off each node
+     individually; back up its metadata folder (for instance, use the following
+     command if your metadata directory is `/var/lib/garage/meta`: `cd
+     /var/lib/garage ; tar -acf meta-v0.9.tar.zst meta/`); turn it back on
+     again.  This will allow you to take a backup of all nodes without
+     impacting global cluster availability.  You can do all nodes of a single
+     zone at once as this does not impact the availability of Garage.
+
+3. Prepare your updated binaries and configuration files for Garage v1.0
+
+4. Shut down all v0.9 nodes simultaneously, and restart them all simultaneously
+   in v1.0.  Use your favorite deployment tool (Ansible, Kubernetes, Nomad) to
+   achieve this as fast as possible.  Garage v1.0 should be in a working state
+   as soon as enough nodes have started.
+
+5. Monitor your cluster in the following hours to see if it works well under
+   your production load.
@@ -0,0 +1,70 @@
+++
+title = "Migrating from 1.0 to 2.0"
+weight = 70
+++
+
+**This guide explains how to migrate to v2.x if you have an existing v1.x.x cluster.
+We don't recommend trying to migrate to v2.x directly from v0.9.x or older.**
+
+This migration procedure has been tested on several clusters without issues.
+However, it is still a *critical procedure* that might cause issues.
+**Make sure to back up all your data before attempting it!**
+
+You might also want to read our [general documentation on upgrading Garage](@/documentation/operations/upgrading.md).
+
+## Changes introduced in v2.0
+
+The following are **breaking changes** in Garage v2.0 that require your attention when migrating:
+
+- The administration API has been completely reworked.
+  Some calls to the `/v1/` endpoints will still work but most will not.
+  New endpoints are prefixed by `/v2/`. **You will need to update all your code that makes use of the admin API.**
+
+- `replication_mode` is no longer a supported configuration parameter,
+  please use `replication_factor` and `consistency_mode` instead.
+
+## Migration procedure
+
+The migration to Garage v2.0 can be done with almost no downtime,
+by restarting all nodes at once in the new version.
+
+The migration steps are as follows:
+
+1. Do a `garage repair --all-nodes --yes tables`, check the logs and check that
+   all data seems to be synced correctly between nodes. If you have time, do
+   additional `garage repair` procedures (`blocks`, `versions`, `block_refs`,
+   etc.)
+
+2. Ensure you have a snapshot of your Garage installation that you can restore
+   to in case the upgrade goes wrong, with one of the following options:
+
+   - You may use the `garage meta snapshot --all` command
+     to make a backup snapshot of the metadata directories of your nodes
+     for backup purposes. Once this command has completed, copy the following
+     files and directories from the `metadata_dir` of all your nodes
+     to somewhere safe: `snapshots`, `cluster_layout`, `data_layout`,
+     `node_key`, `node_key.pub`. (If you have set the `metadata_snapshots_dir`
+     to a different value in your config file, back up that directory instead.)
+
+   - If you are running a filesystem such as ZFS or BTRFS that support
+     snapshotting, you can create a filesystem-level snapshot of the `metadata_dir`
+     of all your nodes to be used as a restoration point if needed.
+
+   - You may also make a back-up manually: turn off each node
+     individually; back up its metadata folder (for instance, use the following
+     command if your metadata directory is `/var/lib/garage/meta`: `cd
+     /var/lib/garage ; tar -acf meta-v1.0.tar.zst meta/`); turn it back on
+     again.  This will allow you to take a backup of all nodes without
+     impacting global cluster availability.  You can do all nodes of a single
+     zone at once as this does not impact the availability of Garage.
+
+3. Prepare your updated binaries and configuration files for Garage v2.0.
+   **Remember to update your configuration file to remove `replication_mode` and replace it by `replication_factor`.**
+
+4. Shut down all v1.0 nodes simultaneously, and restart them all simultaneously
+   in v2.0.  Use your favorite deployment tool (Ansible, Kubernetes, Nomad) to
+   achieve this as fast as possible.  Garage v2.0 should be in a working state
+   as soon as enough nodes have started.
+
+5. Monitor your cluster in the following hours to see if it works well under
+   your production load.
@@ -1,6 +1,6 @@
 +++
 title = "Testing strategy"
-weight = 30
+weight = 100
 +++


@@ -28,16 +28,16 @@ We should try to test in least invasive ways, i.e. minimize the impact of the te
 - Not making `garage` a shared library (launch using `execve`, it's perfectly fine)

 Instead, we should focus on building a clean outer interface for the `garage` binary,
-for example loading configuration using environnement variables instead of the configuration file if that's helpfull for writing the tests.
+for example loading configuration using environment variables instead of the configuration file if that's helpful for writing the tests.

 There are two reasons for this:

- Keep the soure code clean and focused
+- Keep the source code clean and focused
 - Test something that is as close as possible as the true garage that will actually be running

 Reminder: rules of simplicity, concerning changes to Garage's source code.
 Always question what we are doing.
-Never do anything just because it looks nice or because we "think" it might be usefull at some later point but without knowing precisely why/when.
+Never do anything just because it looks nice or because we "think" it might be useful at some later point but without knowing precisely why/when.
 Only do things that make perfect sense in the context of what we currently know.

 ## References
@@ -71,5 +71,3 @@ Interesting blog posts on the blog of the Sled database:
 Misc:
  - [mutagen](https://github.com/llogiq/mutagen) - mutation testing is a way to assert our test quality by mutating the code and see if the mutation makes the tests fail
  - [fuzzing](https://rust-fuzz.github.io/book/) - cargo supports fuzzing, it could be a way to test our software reliability in presence of garbage data.
-  
-
@@ -8,18 +8,22 @@ listen address is specified in the `[admin]` section of the configuration
 file (see [configuration file
 reference](@/documentation/reference-manual/configuration.md))

-**WARNING.** At this point, there is no comittement to stability of the APIs described in this document.
-We will bump the version numbers prefixed to each API endpoint at each time the syntax
-or semantics change, meaning that code that relies on these endpoint will break
+**WARNING.** At this point, there is no commitment to the stability of the APIs described in this document.
+We will bump the version numbers prefixed to each API endpoint each time the syntax
+or semantics change, meaning that code that relies on these endpoints will break
 when changes are introduced.

-The Garage administration API was introduced in version 0.7.2, this document
-does not apply to older versions of Garage.
+The Garage administration API was introduced in version 0.7.2, and was
+changed several times.
+
+**THIS DOCUMENT IS DEPRECATED.** We now have an OpenAPI spec which is automatically generated
+from Garage's source code and is always up-to-date. See `doc/api/garage-admin-v2.html`.
+Text in this document is no longer kept in sync with the admin API's actual behavior.


 ## Access control

-The admin API uses two different tokens for acces control, that are specified in the config file's `[admin]` section:
+The admin API uses two different tokens for access control, that are specified in the config file's `[admin]` section:

 - `metrics_token`: the token for accessing the Metrics endpoint (if this token
  is not set in the config file, the Metrics endpoint can be accessed without
@@ -52,104 +56,132 @@ Returns an HTTP status 200 if the node is ready to answer user's requests,
 and an HTTP status 503 (Service Unavailable) if there are some partitions
 for which a quorum of nodes is not available.
 A simple textual message is also returned in a body with content-type `text/plain`.
-See `/v0/health` for an API that also returns JSON output.
+See `/v2/GetClusterHealth` for an API that also returns JSON output.
+
+### Other special endpoints
+
+#### CheckDomain `GET /check?domain=<domain>`
+
+Checks whether this Garage cluster serves a website for domain `<domain>`.
+Returns HTTP 200 Ok if yes, or HTTP 4xx if no website is available for this domain.

 ### Cluster operations

-#### GetClusterStatus `GET /v0/status`
+#### GetClusterStatus `GET /v2/GetClusterStatus`

 Returns the cluster's current status in JSON, including:

- ID of the node being queried and its version of the Garage daemon
 - Live nodes
 - Currently configured cluster layout
- Staged changes to the cluster layout

 Example response body:

 ```json
 {
-  "node": "ec79480e0ce52ae26fd00c9da684e4fa56658d9c64cdcecb094e936de0bfe71f",
-  "garage_version": "git:v0.8.0",
-  "knownNodes": {
-    "ec79480e0ce52ae26fd00c9da684e4fa56658d9c64cdcecb094e936de0bfe71f": {
-      "addr": "10.0.0.11:3901",
-      "is_up": true,
-      "last_seen_secs_ago": 9,
-      "hostname": "node1"
-    },
-    "4a6ae5a1d0d33bf895f5bb4f0a418b7dc94c47c0dd2eb108d1158f3c8f60b0ff": {
-      "addr": "10.0.0.12:3901",
-      "is_up": true,
-      "last_seen_secs_ago": 1,
-      "hostname": "node2"
-    },
-    "23ffd0cdd375ebff573b20cc5cef38996b51c1a7d6dbcf2c6e619876e507cf27": {
-      "addr": "10.0.0.21:3901",
-      "is_up": true,
-      "last_seen_secs_ago": 7,
-      "hostname": "node3"
-    },
-    "e2ee7984ee65b260682086ec70026165903c86e601a4a5a501c1900afe28d84b": {
-      "addr": "10.0.0.22:3901",
-      "is_up": true,
-      "last_seen_secs_ago": 1,
-      "hostname": "node4"
-    }
-  },
-  "layout": {
-    "version": 12,
-    "roles": {
-      "ec79480e0ce52ae26fd00c9da684e4fa56658d9c64cdcecb094e936de0bfe71f": {
+  "layoutVersion": 5,
+  "nodes": [
+    {
+      "id": "62b218d848e86a64f7fe1909735f29a4350547b54c4b204f91246a14eb0a1a8c",
+      "role": {
+        "id": "62b218d848e86a64f7fe1909735f29a4350547b54c4b204f91246a14eb0a1a8c",
        "zone": "dc1",
-        "capacity": 4,
-        "tags": [
-          "node1"
-        ]
+        "capacity": 100000000000,
+        "tags": []
      },
-      "4a6ae5a1d0d33bf895f5bb4f0a418b7dc94c47c0dd2eb108d1158f3c8f60b0ff": {
-        "zone": "dc1",
-        "capacity": 6,
-        "tags": [
-          "node2"
-        ]
+      "addr": "10.0.0.3:3901",
+      "hostname": "node3",
+      "isUp": true,
+      "lastSeenSecsAgo": 12,
+      "draining": false,
+      "dataPartition": {
+        "available": 660270088192,
+        "total": 873862266880
      },
-      "23ffd0cdd375ebff573b20cc5cef38996b51c1a7d6dbcf2c6e619876e507cf27": {
-        "zone": "dc2",
-        "capacity": 10,
-        "tags": [
-          "node3"
-        ]
+      "metadataPartition": {
+        "available": 660270088192,
+        "total": 873862266880
      }
    },
-    "stagedRoleChanges": {
-      "e2ee7984ee65b260682086ec70026165903c86e601a4a5a501c1900afe28d84b": {
-        "zone": "dc2",
-        "capacity": 5,
-        "tags": [
-          "node4"
-        ]
+    {
+      "id": "a11c7cf18af297379eff8688360155fe68d9061654449ba0ce239252f5a7487f",
+      "role": null,
+      "addr": "10.0.0.2:3901",
+      "hostname": "node2",
+      "isUp": true,
+      "lastSeenSecsAgo": 11,
+      "draining": true,
+      "dataPartition": {
+        "available": 660270088192,
+        "total": 873862266880
+      },
+      "metadataPartition": {
+        "available": 660270088192,
+        "total": 873862266880
+      }
+    },
+    {
+      "id": "a235ac7695e0c54d7b403943025f57504d500fdcc5c3e42c71c5212faca040a2",
+      "role": {
+        "id": "a235ac7695e0c54d7b403943025f57504d500fdcc5c3e42c71c5212faca040a2",
+        "zone": "dc1",
+        "capacity": 100000000000,
+        "tags": []
+      },
+      "addr": "127.0.0.1:3904",
+      "hostname": "lindy",
+      "isUp": true,
+      "lastSeenSecsAgo": 2,
+      "draining": false,
+      "dataPartition": {
+        "available": 660270088192,
+        "total": 873862266880
+      },
+      "metadataPartition": {
+        "available": 660270088192,
+        "total": 873862266880
+      }
+    },
+    {
+      "id": "b10c110e4e854e5aa3f4637681befac755154b20059ec163254ddbfae86b09df",
+      "role": {
+        "id": "b10c110e4e854e5aa3f4637681befac755154b20059ec163254ddbfae86b09df",
+        "zone": "dc1",
+        "capacity": 100000000000,
+        "tags": []
+      },
+      "addr": "10.0.0.1:3901",
+      "hostname": "node1",
+      "isUp": true,
+      "lastSeenSecsAgo": 3,
+      "draining": false,
+      "dataPartition": {
+        "available": 660270088192,
+        "total": 873862266880
+      },
+      "metadataPartition": {
+        "available": 660270088192,
+        "total": 873862266880
      }
    }
-  }
+  ]
 }
 ```

-#### GetClusterHealth `GET /v0/health`
+#### GetClusterHealth `GET /v2/GetClusterHealth`

 Returns the cluster's current health in JSON format, with the following variables:

- `status`: one of `Healthy`, `Degraded` or `Unavailable`:
-  - Healthy: Garage node is connected to all storage nodes
-  - Degraded: Garage node is not connected to all storage nodes, but a quorum of write nodes is available for all partitions
-  - Unavailable: a quorum of write nodes is not available for some partitions
- `known_nodes`: the number of nodes this Garage node has had a TCP connection to since the daemon started
- `connected_nodes`: the nubmer of nodes this Garage node currently has an open connection to
- `storage_nodes`: the number of storage nodes currently registered in the cluster layout
- `storage_nodes_ok`: the number of storage nodes to which a connection is currently open
+- `status`: one of `healthy`, `degraded` or `unavailable`:
+  - healthy: Garage node is connected to all storage nodes
+  - degraded: Garage node is not connected to all storage nodes, but a quorum of write nodes is available for all partitions
+  - unavailable: a quorum of write nodes is not available for some partitions
+- `knownNodes`: the number of nodes this Garage node has had a TCP connection to since the daemon started
+- `connectedNodes`: the number of nodes this Garage node currently has an open connection to
+- `storageNodes`: the number of storage nodes currently registered in the cluster layout
+- `storageNodesOk`: the number of storage nodes to which a connection is currently open
 - `partitions`: the total number of partitions of the data (currently always 256)
- `partitions_quorum`: the number of partitions for which a quorum of write nodes is available
- `partitions_all_ok`: the number of partitions for which we are connected to all storage nodes responsible of storing it
+- `partitionsQuorum`: the number of partitions for which a quorum of write nodes is available
+- `partitionsAllOk`: the number of partitions for which we are connected to all storage nodes responsible of storing it

 Contrarily to `GET /health`, this endpoint always returns a 200 OK HTTP response code.

@@ -157,18 +189,18 @@ Example response body:

 ```json
 {
-    "status": "Degraded",
-    "known_nodes": 3,
-    "connected_nodes": 2,
-    "storage_nodes": 3,
-    "storage_nodes_ok": 2,
-    "partitions": 256,
-    "partitions_quorum": 256,
-    "partitions_all_ok": 0
+  "status": "degraded",
+  "knownNodes": 3,
+  "connectedNodes": 3,
+  "storageNodes": 4,
+  "storageNodesOk": 3,
+  "partitions": 256,
+  "partitionsQuorum": 256,
+  "partitionsAllOk": 64
 }
 ```

-#### ConnectClusterNodes `POST /v0/connect`
+#### ConnectClusterNodes `POST /v2/ConnectClusterNodes`

 Instructs this Garage node to connect to other Garage nodes at specified addresses.

@@ -198,7 +230,7 @@ Example response:
 ]
 ```

-#### GetClusterLayout `GET /v0/layout`
+#### GetClusterLayout `GET /v2/GetClusterLayout`

 Returns the cluster's current layout in JSON, including:

@@ -212,42 +244,54 @@ Example response body:
 ```json
 {
  "version": 12,
-  "roles": {
-    "ec79480e0ce52ae26fd00c9da684e4fa56658d9c64cdcecb094e936de0bfe71f": {
+  "roles": [
+    {
+      "id": "ec79480e0ce52ae26fd00c9da684e4fa56658d9c64cdcecb094e936de0bfe71f",
      "zone": "dc1",
-      "capacity": 4,
+      "capacity": 10737418240,
      "tags": [
        "node1"
      ]
    },
-    "4a6ae5a1d0d33bf895f5bb4f0a418b7dc94c47c0dd2eb108d1158f3c8f60b0ff": {
+    {
+      "id": "4a6ae5a1d0d33bf895f5bb4f0a418b7dc94c47c0dd2eb108d1158f3c8f60b0ff",
      "zone": "dc1",
-      "capacity": 6,
+      "capacity": 10737418240,
      "tags": [
        "node2"
      ]
    },
-    "23ffd0cdd375ebff573b20cc5cef38996b51c1a7d6dbcf2c6e619876e507cf27": {
+    {
+      "id": "23ffd0cdd375ebff573b20cc5cef38996b51c1a7d6dbcf2c6e619876e507cf27",
      "zone": "dc2",
-      "capacity": 10,
+      "capacity": 10737418240,
      "tags": [
        "node3"
      ]
    }
-  },
-  "stagedRoleChanges": {
-    "e2ee7984ee65b260682086ec70026165903c86e601a4a5a501c1900afe28d84b": {
+  ],
+  "stagedRoleChanges": [
+    {
+      "id": "e2ee7984ee65b260682086ec70026165903c86e601a4a5a501c1900afe28d84b",
+      "remove": false,
      "zone": "dc2",
-      "capacity": 5,
+      "capacity": 10737418240,
      "tags": [
        "node4"
      ]
    }
-  }
+    {
+      "id": "23ffd0cdd375ebff573b20cc5cef38996b51c1a7d6dbcf2c6e619876e507cf27",
+      "remove": true,
+      "zone": null,
+      "capacity": null,
+      "tags": null,
+    }
+  ]
 }
 ```

-#### UpdateClusterLayout `POST /v0/layout`
+#### UpdateClusterLayout `POST /v2/UpdateClusterLayout`

 Send modifications to the cluster layout. These modifications will
 be included in the staged role changes, visible in subsequent calls
@@ -259,8 +303,9 @@ the layout.
 Request body format:

 ```json
-{
-  <node_id>: {
+[
+  {
+    "id": <node_id>,
    "capacity": <new_capacity>,
    "zone": <new_zone>,
    "tags": [
@@ -268,17 +313,22 @@ Request body format:
      ...
    ]
  },
-  <node_id_to_remove>: null,
-  ...
-}
+  {
+    "id": <node_id_to_remove>,
+    "remove": true
+  }
+]
 ```

 Contrary to the CLI that may update only a subset of the fields
 `capacity`, `zone` and `tags`, when calling this API all of these
 values must be specified.

+This returns the new cluster layout with the proposed staged changes,
+as returned by GetClusterLayout.

-#### ApplyClusterLayout `POST /v0/layout/apply`
+
+#### ApplyClusterLayout `POST /v2/ApplyClusterLayout`

 Applies to the cluster the layout changes currently registered as
 staged layout changes.
@@ -295,28 +345,22 @@ Similarly to the CLI, the body must include the version of the new layout
 that will be created, which MUST be 1 + the value of the currently
 existing layout in the cluster.

-#### RevertClusterLayout `POST /v0/layout/revert`
+This returns the message describing all the calculations done to compute the new
+layout, as well as the description of the layout as returned by GetClusterLayout.
+
+#### RevertClusterLayout `POST /v2/RevertClusterLayout`

 Clears all of the staged layout changes.

-Request body format:
+This requests contains an empty body.

-```json
-{
-  "version": 13
-}
-```
-
-Reverting the staged changes is done by incrementing the version number
-and clearing the contents of the staged change list.
-Similarly to the CLI, the body must include the incremented
-version number, which MUST be 1 + the value of the currently
-existing layout in the cluster.
+This returns the new cluster layout with all changes reverted,
+as returned by GetClusterLayout.


 ### Access key operations

-#### ListKeys `GET /v0/key`
+#### ListKeys `GET /v2/ListKeys`

 Returns all API access keys in the cluster.

@@ -335,34 +379,8 @@ Example response:
 ]
 ```

-#### CreateKey `POST /v0/key`
-
-Creates a new API access key.
-
-Request body format:
-
-```json
-{
-    "name": "NameOfMyKey"
-}
-```
-
-#### ImportKey `POST /v0/key/import`
-
-Imports an existing API key.
-
-Request body format:
-
-```json
-{
-    "accessKeyId": "GK31c2f218a2e44f485b94239e",
-    "secretAccessKey": "b892c0665f0ada8a4755dae98baa3b133590e11dae3bcc1f9d769d67f16c3835",
-    "name": "NameOfMyKey"
-}
-```
-
-#### GetKeyInfo `GET /v0/key?id=<acces key id>`
-#### GetKeyInfo `GET /v0/key?search=<pattern>`
+#### GetKeyInfo `GET /v2/GetKeyInfo?id=<access key id>`
+#### GetKeyInfo `GET /v2/GetKeyInfo?search=<pattern>`

 Returns information about the requested API access key.

@@ -370,6 +388,9 @@ If `id` is set, the key is looked up using its exact identifier (faster).
 If `search` is set, the key is looked up using its name or prefix
 of identifier (slower, all keys are enumerated to do this).

+Optionally, the query parameter `showSecretKey=true` can be set to reveal the
+associated secret access key.
+
 Example response:

 ```json
@@ -433,11 +454,40 @@ Example response:
 }
 ```

-#### DeleteKey `DELETE /v0/key?id=<acces key id>`
+#### CreateKey `POST /v2/CreateKey`

-Deletes an API access key.
+Creates a new API access key.

-#### UpdateKey `POST /v0/key?id=<acces key id>`
+Request body format:
+
+```json
+{
+    "name": "NameOfMyKey"
+}
+```
+
+This returns the key info, including the created secret key,
+in the same format as the result of GetKeyInfo.
+
+#### ImportKey `POST /v2/ImportKey`
+
+Imports an existing API key.
+This will check that the imported key is in the valid format, i.e.
+is a key that could have been generated by Garage.
+
+Request body format:
+
+```json
+{
+    "accessKeyId": "GK31c2f218a2e44f485b94239e",
+    "secretAccessKey": "b892c0665f0ada8a4755dae98baa3b133590e11dae3bcc1f9d769d67f16c3835",
+    "name": "NameOfMyKey"
+}
+```
+
+This returns the key info in the same format as the result of GetKeyInfo.
+
+#### UpdateKey `POST /v2/UpdateKey?id=<access key id>`

 Updates information about the specified API access key.

@@ -457,10 +507,16 @@ All fields (`name`, `allow` and `deny`) are optional.
 If they are present, the corresponding modifications are applied to the key, otherwise nothing is changed.
 The possible flags in `allow` and `deny` are: `createBucket`.

+This returns the key info in the same format as the result of GetKeyInfo.
+
+#### DeleteKey `POST /v2/DeleteKey?id=<access key id>`
+
+Deletes an API access key.
+

 ### Bucket operations

-#### ListBuckets `GET /v0/bucket`
+#### ListBuckets `GET /v2/ListBuckets`

 Returns all storage buckets in the cluster.

@@ -502,8 +558,8 @@ Example response:
 ]
 ```

-#### GetBucketInfo `GET /v0/bucket?id=<bucket id>`
-#### GetBucketInfo `GET /v0/bucket?globalAlias=<alias>`
+#### GetBucketInfo `GET /v2/GetBucketInfo?id=<bucket id>`
+#### GetBucketInfo `GET /v2/GetBucketInfo?globalAlias=<alias>`

 Returns information about the requested storage bucket.

@@ -535,7 +591,10 @@ Example response:
        ],
        "objects": 14827,
        "bytes": 13189855625,
-        "unfinshedUploads": 0,
+        "unfinishedUploads": 1,
+        "unfinishedMultipartUploads": 1,
+        "unfinishedMultipartUploadParts": 11,
+        "unfinishedMultipartUploadBytes": 41943040,
        "quotas": {
            "maxSize": null,
            "maxObjects": null
@@ -543,7 +602,7 @@ Example response:
 }
 ```

-#### CreateBucket `POST /v0/bucket`
+#### CreateBucket `POST /v2/CreateBucket`

 Creates a new storage bucket.

@@ -583,13 +642,7 @@ or no alias at all.
 Technically, you can also specify both `globalAlias` and `localAlias` and that would create
 two aliases, but I don't see why you would want to do that.

-#### DeleteBucket `DELETE /v0/bucket?id=<bucket id>`
-
-Deletes a storage bucket. A bucket cannot be deleted if it is not empty.
-
-Warning: this will delete all aliases associated with the bucket!
-
-#### UpdateBucket `PUT /v0/bucket?id=<bucket id>`
+#### UpdateBucket `POST /v2/UpdateBucket?id=<bucket id>`

 Updates configuration of the given bucket.

@@ -621,9 +674,38 @@ In `quotas`: new values of `maxSize` and `maxObjects` must both be specified, or
 to remove the quotas. An absent value will be considered the same as a `null`. It is not possible
 to change only one of the two quotas.

+#### DeleteBucket `POST /v2/DeleteBucket?id=<bucket id>`
+
+Deletes a storage bucket. A bucket cannot be deleted if it is not empty.
+
+Warning: this will delete all aliases associated with the bucket!
+
+#### CleanupIncompleteUploads `POST /v2/CleanupIncompleteUploads`
+
+Cleanup all incomplete uploads in a bucket that are older than a specified number
+of seconds.
+
+Request body format:
+
+```json
+{
+    "bucketId": "e6a14cd6a27f48684579ec6b381c078ab11697e6bc8513b72b2f5307e25fff9b",
+    "olderThanSecs": 3600
+}
+```
+
+Response format
+
+```json
+{
+    "uploadsDeleted": 12
+}
+```
+
+
 ### Operations on permissions for keys on buckets

-#### BucketAllowKey `POST /v0/bucket/allow`
+#### AllowBucketKey `POST /v2/AllowBucketKey`

 Allows a key to do read/write/owner operations on a bucket.

@@ -644,7 +726,7 @@ Request body format:
 Flags in `permissions` which have the value `true` will be activated.
 Other flags will remain unchanged.

-#### BucketDenyKey `POST /v0/bucket/deny`
+#### DenyBucketKey `POST /v2/DenyBucketKey`

 Denies a key from doing read/write/owner operations on a bucket.

@@ -668,19 +750,35 @@ Other flags will remain unchanged.

 ### Operations on bucket aliases

-#### GlobalAliasBucket `PUT /v0/bucket/alias/global?id=<bucket id>&alias=<global alias>`
+#### AddBucketAlias `POST /v2/AddBucketAlias`

-Empty body. Creates a global alias for a bucket.
+Creates an alias for a bucket in the namespace of a specific access key.
+To create a global alias, specify the `globalAlias` field.
+To create a local alias, specify the `localAlias` and `accessKeyId` fields.

-#### GlobalUnaliasBucket `DELETE /v0/bucket/alias/global?id=<bucket id>&alias=<global alias>`
+Request body format:

-Removes a global alias for a bucket.
+```json
+{
+    "bucketId": "e6a14cd6a27f48684579ec6b381c078ab11697e6bc8513b72b2f5307e25fff9b",
+    "globalAlias": "my-bucket"
+}
+```

-#### LocalAliasBucket `PUT /v0/bucket/alias/local?id=<bucket id>&accessKeyId=<access key ID>&alias=<local alias>`
+or:

-Empty body. Creates a local alias for a bucket in the namespace of a specific access key.
+```json
+{
+    "bucketId": "e6a14cd6a27f48684579ec6b381c078ab11697e6bc8513b72b2f5307e25fff9b",
+    "accessKeyId": "GK31c2f218a2e44f485b94239e",
+    "localAlias": "my-bucket"
+}
+```

-#### LocalUnaliasBucket `DELETE /v0/bucket/alias/local?id=<bucket id>&accessKeyId<access key ID>&alias=<local alias>`
+#### RemoveBucketAlias `POST /v2/RemoveBucketAlias`

-Removes a local alias for a bucket in the namespace of a specific access key.
+Removes an alias for a bucket in the namespace of a specific access key.
+To remove a global alias, specify the `globalAlias` field.
+To remove a local alias, specify the `localAlias` and `accessKeyId` fields.

+Request body format: same as AddBucketAlias.
@@ -35,7 +35,7 @@ Triples in K2V are constituted of three fields:
  partition key in which the client wants to read/delete lists of items

 - a sort key (`sk`), an utf8 string that defines the index of the triplet inside its
-  partition; triplets are uniquely idendified by their partition key + sort key
+  partition; triplets are uniquely identified by their partition key + sort key

 - a value (`v`), an opaque binary blob associated to the partition key + sort key;
  they are transmitted as binary when possible but in most case in the JSON API
@@ -74,7 +74,7 @@ are obsoleted by the new write.

 **Basic insertion.** To insert a new value `v4` with context `[(node1, t2), (node2, t3)]`, in a
 simple case where there was no insertion in-between reading the value
-mentionned above and writing `v4`, and supposing that node2 receives the
+mentioned above and writing `v4`, and supposing that node2 receives the
 InsertItem query:

 - `node2` generates a timestamp `t4` such that `t4 > t3`.
@@ -146,7 +146,7 @@ in a bucket, as the partition key becomes the sort key in the index.
 How indexing works:

 - Each node keeps a local count of how many items it stores for each partition,
-  in a local Sled tree that is updated atomically when an item is modified.
+  in a local database tree that is updated atomically when an item is modified.
 - These local counters are asynchronously stored in the index table which is
  a regular Garage table spread in the network. Counters are stored as LWW values,
  so basically the final table will have the following structure:
@@ -332,7 +332,7 @@ Inserts a single item. This request does not use JSON, the body is sent directly

 To supersede previous values, the HTTP header `X-Garage-Causality-Token` should
 be set to the causality token returned by a previous read on this key. This
-header can be ommitted for the first writes to the key.
+header can be omitted for the first writes to the key.

 Example query:

@@ -397,7 +397,7 @@ smallest partition key that exists.  It returns partition keys in increasing
 order, or decreasing order if `reverse` is set to `true`,
 and stops when either of the following conditions is met:

-1. if `end` is specfied, the partition key `end` is reached or surpassed (if it
+1. if `end` is specified, the partition key `end` is reached or surpassed (if it
   is reached exactly, it is not included in the result)

 2. if `limit` is specified, `limit` partition keys have been listed
@@ -491,7 +491,7 @@ the triplet is inserted for the first time, the causality token should be set to

 The value is expected to be a base64-encoded binary blob. The value `null` can
 also be used to delete the triplet while preserving causality information: this
-allows to know if a delete has happenned concurrently with an insert, in which
+allows to know if a delete has happened concurrently with an insert, in which
 case both are preserved and returned on reads (see below).

 Partition keys and sort keys are utf8 strings which are stored sorted by
@@ -540,7 +540,7 @@ JSON struct with the following fields:

 For each of the searches, triplets are listed and returned separately. The
 semantics of `prefix`, `start`, `end`, `limit` and `reverse` are the same as for ReadIndex. The
-additionnal parameter `singleItem` allows to get a single item, whose sort key
+additional parameter `singleItem` allows to get a single item, whose sort key
 is the one given in `start`. Parameters `conflictsOnly` and `tombstones`
 control additional filters on the items that are returned.

@@ -562,7 +562,7 @@ token>", v: ["<value1>", ...] }`, with the following fields:
 - in case of concurrent update and deletion, a `null` is added to the list of concurrent values

 - if the `tombstones` query parameter is set to `true`, tombstones are returned
-  for items that have been deleted (this can be usefull for inserting after an
+  for items that have been deleted (this can be useful for inserting after an
  item that has been deleted, so that the insert is not considered
  concurrent with the delete). Tombstones are returned as tuples in the
  same format with only `null` values
@@ -0,0 +1,13 @@
+optimal_layout.aux
+optimal_layout.log
+optimal_layout.synctex.gz
+optimal_layout.bbl
+optimal_layout.blg
+
+geodistrib.aux
+geodistrib.bbl
+geodistrib.blg
+geodistrib.log
+geodistrib.out
+geodistrib.synctex.gz
+
@@ -0,0 +1,317 @@
+\documentclass[]{article}
+
+\usepackage{amsmath,amssymb}
+\usepackage{amsthm}
+
+\usepackage{stmaryrd}
+
+\usepackage{graphicx,xcolor}
+\usepackage{hyperref}
+
+\usepackage{algorithm,algpseudocode,float} 
+
+\renewcommand\thesubsubsection{\Alph{subsubsection})}
+
+\newtheorem{proposition}{Proposition}
+
+%opening
+\title{An algorithm for geo-distributed and redundant storage in Garage}
+\author{Mendes Oulamara \\ \emph{mendes@deuxfleurs.fr}}
+\date{}
+
+\begin{document}
+
+\maketitle
+
+\begin{abstract}
+	Garage
+\end{abstract}
+
+\section{Introduction}
+
+Garage\footnote{\url{https://garagehq.deuxfleurs.fr/}} is an open-source distributed object storage service tailored for self-hosting. It was designed by the Deuxfleurs association\footnote{\url{https://deuxfleurs.fr/}} to enable small structures (associations, collectives, small companies) to share storage resources to reliably self-host their data, possibly with old and non-reliable machines.
+
+To achieve these reliability and availability goals, the data is broken into \emph{partitions} and every partition is replicated over 3 different machines (that we call \emph{nodes}). When the data is queried, a consensus algorithm allows to fetch it from one of the nodes. A \emph{replication factor} of 3 ensures the best guarantees in the consensus algorithm \cite{ADD RREF}, but this parameter can be different.
+
+Moreover, if the nodes are spread over different \emph{zones} (different houses, offices, cities\dots), we can ask the data to be replicated over nodes belonging to different zones, to improve the storage robustness against zone failure (such as power outage). To do so, we set a \emph{redundancy parameter}, that is no more than the replication factor, and we ask that any partition is replicated over this number of zones at least.
+
+In this work, we propose a repartition algorithm that, given the nodes specifications and the replication and redundancy parameters, computes an optimal assignation of partitions to nodes. We say that the assignation is optimal in the sense that it maximizes the size of the partitions, and hence the effective storage capacity of the system.
+
+Moreover, when a former assignation exists, which is not optimal anymore due to nodes or zones updates, our algorithm computes a new optimal assignation that minimizes the amount of data to be transferred during the assignation update (the \emph{transfer load}).
+
+We call the set of nodes cooperating to store the data a \emph{cluster}, and a description of the nodes, zones and the assignation of partitions to nodes a \emph{cluster layout}
+
+\subsection{Notations}
+
+Let $k$ be some fixed parameter value, typically 8, that we call the ``partition bits''.
+Every object to be stored in the system is split into data blocks of fixed size. We compute a hash $h(\mathbf{b})$ of every such block $\mathbf{b}$, and we define the $k$ last bits of this hash to be the partition number $p(\mathbf{b})$ of the block. This label can take $P=2^k$ different values, and hence there are $P$ different partitions. We denote $\mathbf{P}$ the set of partition labels (i.e. $\mathbf{P}=\llbracket1,P\rrbracket$).
+
+We are given a set $\mathbf{N}$ of $N$ nodes  and a set  $\mathbf{Z}$ of $Z$ zones. Every node $n$ has a non-negative storage capacity $c_n\ge 0$ and belongs to a zone $z_n\in \mathbf{Z}$. We are also given a replication parameter $\rho_\mathbf{N}$ and a redundancy parameter $\rho_\mathbf{Z}$ such that  $1\le \rho_\mathbf{Z} \le \rho_\mathbf{N}$ (typical values would be $\rho_N=3$ and $\rho_Z=2$).
+
+Our goal is to compute an assignment $\alpha = (\alpha_p^1, \ldots, \alpha_p^{\rho_\mathbf{N}})_{p\in \mathbf{P}}$ such that every partition $p$ is associated to $\rho_\mathbf{N}$ distinct nodes $\alpha_p^1, \ldots, \alpha_p^{\rho_\mathbf{N}} \in \mathbf{N}$ and these nodes belong to at least $\rho_\mathbf{Z}$ distinct zones. Among the possible assignations, we choose one that \emph{maximizes} the effective storage capacity of the cluster. If the layout contained a previous assignment $\alpha'$, we \emph{minimize} the amount of data to transfer during the layout update by making $\alpha$ as close as possible to $\alpha'$. These maximization and minimization are described more formally in the following section.
+
+\subsection{Optimization parameters}
+
+To link the effective storage capacity of the cluster to partition assignment, we make the following assumption:
+\begin{equation}
+	\tag{H1}
+	\text{\emph{All partitions have the same size $s$.}}
+\end{equation}
+This assumption is justified by the dispersion of the hashing function, when the number of partitions is small relative to the number of stored blocks.
+
+Every node $n$ will store some number $p_n$ of partitions (it is the number of partitions $p$ such that $n$ appears in the $\alpha_p$). Hence the partitions stored by $n$ (and hence all partitions by our assumption) have there size bounded by $c_n/p_n$. This remark leads us to define the optimal size that we will want to maximize:
+
+\begin{equation}
+	\label{eq:optimal}
+	\tag{OPT}
+s^* = \min_{n \in N} \frac{c_n}{p_n}.
+\end{equation}
+
+When the capacities of the nodes are updated (this includes adding or removing a node), we want to update the assignment as well. However, transferring the data between nodes has a cost and we would like to limit the number of changes in the assignment. We make the following assumption:
+\begin{equation}
+	\tag{H2}
+	\text{\emph{Nodes updates happen rarely relatively to block operations.}}
+\end{equation}
+This assumption justifies that when we compute the new assignment $\alpha$, it is worth to optimize the partition size \eqref{eq:optimal} first, and then, among the possible optimal solution, to try to minimize the number of partition transfers. More formally, we minimize the distance between two assignments defined by
+\begin{equation}
+	d(\alpha, \alpha') := \#\{ (n,p) \in \mathbf{N}\times\mathbf{P} ~|~ n\in \alpha_p \triangle \alpha'_p \}
+\end{equation}
+where the symmetric difference $\alpha_p \triangle \alpha'_p$ denotes the nodes appearing in one of the assignations but not in both.
+
+\section{Computation of an optimal assignment}
+
+The algorithm that we propose takes as inputs the cluster layout parameters $\mathbf{N}$, $\mathbf{Z}$, $\mathbf{P}$, $(c_n)_{n\in \mathbf{N}}$, $\rho_\mathbf{N}$, $\rho_\mathbf{Z}$, that we defined in the introduction, together with the former assignation $\alpha'$ (if any). The computation of the new optimal assignation $\alpha^*$ is done in three successive steps that will be detailed in the following sections. The first step computes the largest partition size $s^*$ that an assignation can achieve. The second step computes an optimal candidate assignment $\alpha$ that achieves $s^*$ and a heuristic is used in the computation to make it hopefully close to $\alpha'$. The third steps modifies $\alpha$ iteratively to reduces $d(\alpha, \alpha')$ and yields an assignation $\alpha^*$ achieving $s^*$, and minimizing $d(\cdot, \alpha')$ among such assignations.
+
+We will explain in the next section how to represent an assignment $\alpha$ by a flow $f$ on a weighted graph $G$ to enable the use of flow and graph algorithms. The main function of the algorithm can be written as follows.
+
+\subsubsection*{Algorithm}
+
+\begin{algorithmic}[1]
+	\Function{Compute Layout}{$\mathbf{N}$, $\mathbf{Z}$, $\mathbf{P}$, $(c_n)_{n\in \mathbf{N}}$, $\rho_\mathbf{N}$, $\rho_\mathbf{Z}$, $\alpha'$}
+	\State $s^* \leftarrow$ \Call{Compute Partition Size}{$\mathbf{N}$, $\mathbf{Z}$, $\mathbf{P}$, $(c_n)_{n\in \mathbf{N}}$, $\rho_\mathbf{N}$, $\rho_\mathbf{Z}$}
+	\State $G \leftarrow G(s^*)$
+	\State $f \leftarrow$ 	\Call{Compute Candidate Assignment}{$G$, $\alpha'$}
+	\State $f^* \leftarrow$ 	\Call{Minimize transfer load}{$G$, $f$, $\alpha'$}
+	\State Build $\alpha^*$ from $f^*$
+	\State \Return $\alpha^*$
+	\EndFunction
+\end{algorithmic}
+
+\subsubsection*{Complexity}
+As we will see in the next sections, the worst case complexity of this algorithm is $O(P^2 N^2)$. The minimization of transfer load is the most expensive step, and it can run with a timeout since it is only an optimization step. Without this step (or with a smart timeout), the worst cas complexity can be $O((PN)^{3/2}\log C)$ where $C$ is the total storage capacity of the cluster. 
+
+\subsection{Determination of the partition size $s^*$}
+
+We will represent an assignment $\alpha$ as a flow in a specific graph $G$. We will not compute the optimal partition size $s^*$ a priori, but we will determine it by dichotomy, as the largest size $s$ such that the maximal flow achievable on $G=G(s)$ has value $\rho_\mathbf{N}P$. We will assume that the capacities are given in a small enough unit (say, Megabytes), and we will determine $s^*$ at the precision of the given unit.
+
+Given some candidate size value $s$, we describe the oriented weighted graph $G=(V,E)$ with vertex set $V$ arc set $E$ (see Figure \ref{fig:flowgraph}).
+
+The set of vertices $V$ contains the source $\mathbf{s}$, the sink $\mathbf{t}$, vertices 
+$\mathbf{p^+, p^-}$ for every partition $p$, vertices $\mathbf{x}_{p,z}$ for every partition $p$ and zone $z$, and vertices $\mathbf{n}$ for every node $n$. 
+
+The set of arcs $E$ contains:
+\begin{itemize}
+	\item ($\mathbf{s}$,$\mathbf{p}^+$, $\rho_\mathbf{Z}$) for every partition $p$;
+	\item ($\mathbf{s}$,$\mathbf{p}^-$, $\rho_\mathbf{N}-\rho_\mathbf{Z}$) for every partition $p$;
+	\item ($\mathbf{p}^+$,$\mathbf{x}_{p,z}$, 1) for every partition $p$ and zone $z$;
+	\item ($\mathbf{p}^-$,$\mathbf{x}_{p,z}$, $\rho_\mathbf{N}-\rho_\mathbf{Z}$) for every partition $p$ and zone $z$;
+	\item ($\mathbf{x}_{p,z}$,$\mathbf{n}$, 1) for every partition $p$, zone $z$ and node $n\in z$;
+	\item ($\mathbf{n}$, $\mathbf{t}$, $\lfloor c_n/s \rfloor$) for every node $n$.
+\end{itemize}
+
+\begin{figure}
+	\centering
+	\includegraphics[width=\linewidth]{figures/flow_graph_param}
+	\caption{An example of graph $G(s)$. Arcs are oriented from left to right, and unlabeled arcs have capacity 1. In this example, nodes $n_1,n_2,n_3$ belong to zone $z_1$, and nodes $n_4,n_5$ belong to zone $z_2$.}
+	\label{fig:flowgraph}
+\end{figure}
+
+In the following complexity calculations, we will use the number of vertices and edges of $G$. Remark from now that $\# V = O(PZ)$ and $\# E = O(PN)$.
+
+\begin{proposition}
+	An assignment $\alpha$ is realizable with partition size $s$ and the redundancy constraints $(\rho_\mathbf{N},\rho_\mathbf{Z})$ if and only if there exists a maximal flow function $f$ in $G$ with total flow $\rho_\mathbf{N}P$, such that the arcs ($\mathbf{x}_{p,z}$,$\mathbf{n}$, 1) used are exactly those for which $p$ is associated to $n$ in $\alpha$.
+\end{proposition}
+\begin{proof}
+	Given such flow $f$, we can reconstruct a candidate $\alpha$. In $f$, the flow passing through $\mathbf{p^+}$ and $\mathbf{p^-}$ is $\rho_\mathbf{N}$, and since the outgoing capacity of every $\mathbf{x}_{p,z}$ is 1, every partition is associated to $\rho_\mathbf{N}$ distinct nodes. The fraction $\rho_\mathbf{Z}$ of the flow passing through every $\mathbf{p^+}$ must be spread over as many distinct zones as every arc outgoing from $\mathbf{p^+}$ has capacity 1. So the reconstructed $\alpha$ verifies the redundancy constraints. For every node $n$, the flow between $\mathbf{n}$ and $\mathbf{t}$ corresponds to the number of partitions associated to $n$. By construction of $f$, this does not exceed $\lfloor c_n/s \rfloor$. We assumed that the partition size is $s$, hence this association does not exceed the storage capacity of the nodes.
+	
+	In the other direction, given an assignment $\alpha$, one can similarly check that the facts that $\alpha$ respects the redundancy constraints, and the storage capacities of the nodes, are necessary condition to construct a maximal flow function $f$.
+\end{proof}
+
+\textbf{Implementation remark:} In the flow algorithm, while exploring the graph, we explore the neighbours of every vertex in a random order to heuristically spread the associations between nodes and partitions.
+
+\subsubsection*{Algorithm}
+With this result mind, we can describe the first step of our algorithm. All divisions are supposed to be integer divisions.
+\begin{algorithmic}[1]
+	\Function{Compute Partition Size}{$\mathbf{N}$, $\mathbf{Z}$, $\mathbf{P}$, $(c_n)_{n\in \mathbf{N}}$, $\rho_\mathbf{N}$, $\rho_\mathbf{Z}$}
+	
+	\State Build the graph $G=G(s=1)$
+	\State $ f \leftarrow$ \Call{Maximal flow}{$G$}
+	\If{$f.\mathrm{total flow} < \rho_\mathbf{N}P$}
+	
+	\State \Return Error: capacities too small or constraints too strong.
+	\EndIf
+	
+	\State $s^- \leftarrow 1$
+	\State $s^+ \leftarrow 1+\frac{1}{\rho_\mathbf{N}}\sum_{n \in \mathbf{N}} c_n$
+	
+	\While{$s^-+1 < s^+$}
+	\State Build the graph $G=G(s=(s^-+s^+)/2)$
+	\State $ f \leftarrow$ \Call{Maximal flow}{$G$}
+	\If{$f.\mathrm{total flow} < \rho_\mathbf{N}P$}
+	\State $s^+ \leftarrow (s^- + s^+)/2$
+	\Else
+	\State $s^- \leftarrow (s^- + s^+)/2$
+	\EndIf
+	\EndWhile
+	
+	\State \Return $s^-$
+	\EndFunction
+\end{algorithmic}
+
+\subsubsection*{Complexity}
+
+To compute the maximal flow, we use Dinic's algorithm. Its complexity on general graphs is $O(\#V^2 \#E)$, but on graphs with edge capacity bounded by a constant, it turns out to be $O(\#E^{3/2})$. The graph $G$ does not fall in this case since the capacities of the arcs incoming to $\mathbf{t}$ are far from bounded. However, the proof of this complexity function works readily for graphs where we only ask the edges \emph{not} incoming to the sink $\mathbf{t}$ to have their capacities bounded by a constant. One can find the proof of this claim in \cite[Section 2]{even1975network}.
+The dichotomy adds a logarithmic factor $\log (C)$ where $C=\sum_{n \in \mathbf{N}} c_n$ is the total capacity of the cluster. The total complexity of this first function is hence 
+$O(\#E^{3/2}\log C ) = O\big((PN)^{3/2} \log C\big)$.
+
+\subsubsection*{Metrics}
+We can display the discrepancy between the computed $s^*$ and the best size we could have hoped for the given total capacity, that is $C/\rho_\mathbf{N}$.
+
+\subsection{Computation of a candidate assignment}
+
+Now that we have the optimal partition size $s^*$, to compute a candidate assignment it would be enough to compute a maximal flow function $f$ on $G(s^*)$. This is what we do if there is no former assignation $\alpha'$.
+
+If there is some $\alpha'$, we add a step that will heuristically help to obtain a candidate $\alpha$ closer to $\alpha'$. We fist compute a flow function $\tilde{f}$ that uses only the partition-to-node associations appearing in $\alpha'$. Most likely, $\tilde{f}$ will not be a maximal flow of $G(s^*)$. In Dinic's algorithm, we can start from a non maximal flow function and then discover improving paths. This is what we do by starting from $\tilde{f}$. The hope\footnote{This is only a hope, because one can find examples where the construction of $f$ from $\tilde{f}$ produces an assignment $\alpha$ that is not as close as possible to $\alpha'$.} is that the final flow function $f$ will tend to keep the associations appearing in $\tilde{f}$.
+
+More formally, we construct the graph $G_{|\alpha'}$ from $G$ by removing all the arcs $(\mathbf{x}_{p,z},\mathbf{n}, 1)$ where $p$ is not associated to $n$ in $\alpha'$. We compute a maximal flow function $\tilde{f}$ in $G_{|\alpha'}$. The flow $\tilde{f}$ is also a valid (most likely non maximal) flow function on $G$. We compute a maximal flow function $f$ on $G$ by starting Dinic's algorithm on $\tilde{f}$.
+
+\subsubsection*{Algorithm}
+\begin{algorithmic}[1]
+	\Function{Compute Candidate Assignment}{$G$, $\alpha'$}
+	\State Build the graph $G_{|\alpha'}$
+	\State $ \tilde{f} \leftarrow$ \Call{Maximal flow}{$G_{|\alpha'}$}
+	\State $ f \leftarrow$ \Call{Maximal flow from flow}{$G$, $\tilde{f}$}
+	\State \Return $f$
+	\EndFunction
+\end{algorithmic}
+
+~ 
+
+\textbf{Remark:} The function ``Maximal flow'' can be just seen as the function ``Maximal flow from flow'' called with the zero flow function as starting flow.
+
+\subsubsection*{Complexity}
+With the considerations of the last section, we have the complexity of the Dinic's algorithm $O(\#E^{3/2}) = O((PN)^{3/2})$.
+
+\subsubsection*{Metrics}
+
+We can display the flow value of $\tilde{f}$, which is an upper bound of the distance between $\alpha$ and $\alpha'$. It might be more a Debug level display than Info.
+
+\subsection{Minimization of the transfer load}
+
+Now that we have a candidate flow function $f$, we want to modify it to make its corresponding assignation $\alpha$ as close as possible to $\alpha'$. Denote by $f'$ the maximal flow corresponding to $\alpha'$, and let $d(f, \alpha')=d(f, f'):=d(\alpha,\alpha')$\footnote{It is the number of arcs of type $(\mathbf{x}_{p,z},\mathbf{n})$ saturated in one flow and not in the other.}. 
+We want to build a sequence $f=f_0, f_1, f_2 \dots$ of maximal flows such that $d(f_i, \alpha')$ decreases as $i$ increases. The distance being a non-negative integer, this sequence of flow functions must be finite. We now explain how to find some improving $f_{i+1}$ from $f_i$.
+
+For any maximal flow $f$ in $G$, we define the oriented weighted graph $G_f=(V, E_f)$ as follows. The vertices of $G_f$ are the same as the vertices of $G$. $E_f$ contains the arc $(v_1,v_2, w)$ between vertices $v_1,v_2\in V$ with weight $w$ if and only if the arc $(v_1,v_2)$ is not saturated in $f$ (i.e. $c(v_1,v_2)-f(v_1,v_2) \ge 1$, we also consider reversed arcs). The weight $w$ is: 
+\begin{itemize}
+	\item  $-1$  if $(v_1,v_2)$ is of type $(\mathbf{x}_{p,z},\mathbf{n})$ or $(\mathbf{x}_{p,z},\mathbf{n})$ and is saturated in only one of the two flows $f,f'$;
+	\item $+1$  if $(v_1,v_2)$ is of type $(\mathbf{x}_{p,z},\mathbf{n})$ or $(\mathbf{x}_{p,z},\mathbf{n})$ and is saturated in either both or none of the two flows $f,f'$;
+	\item $0$ otherwise.
+\end{itemize}
+
+If $\gamma$ is a simple cycle of arcs in $G_f$, we define its weight $w(\gamma)$ as the sum of the weights of its arcs. We can add $+1$ to the value of $f$ on the arcs of $\gamma$, and by construction of $G_f$ and the fact that $\gamma$ is a cycle, the function that we get is still a valid flow function on $G$, it is maximal as it has the same flow value as $f$. We denote this new function $f+\gamma$.
+
+\begin{proposition}
+	Given a maximal flow $f$ and a simple cycle $\gamma$ in $G_f$, we have $d(f+\gamma, f') - d(f,f') = w(\gamma)$.
+\end{proposition}
+\begin{proof}
+	Let $X$ be the set of arcs of type $(\mathbf{x}_{p,z},\mathbf{n})$. Then we can express $d(f,f')$ as
+	\begin{align*}
+		d(f,f') & = \#\{e\in X ~|~ f(e)\neq f'(e)\}
+		= \sum_{e\in X} 1_{f(e)\neq f'(e)} \\
+		& = \frac{1}{2}\big( \#X + \sum_{e\in X} 1_{f(e)\neq f'(e)} - 1_{f(e)= f'(e)} \big).
+	\end{align*}
+	We can express the cycle weight as
+	\begin{align*}
+		w(\gamma) &  = \sum_{e\in X, e\in \gamma} - 1_{f(e)\neq f'(e)} + 1_{f(e)= f'(e)}.
+	\end{align*}
+	Remark that since we passed on unit of flow in $\gamma$ to construct $f+\gamma$, we have for any $e\in X$, $f(e)=f'(e)$ if and only if $(f+\gamma)(e) \neq f'(e)$.
+	Hence
+	\begin{align*}
+		w(\gamma) &  = \frac{1}{2}(w(\gamma) + w(\gamma)) \\
+		&= \frac{1}{2} \Big( 
+		\sum_{e\in X, e\in \gamma} - 1_{f(e)\neq f'(e)} +  1_{f(e)= f'(e)} \\
+		& \qquad +
+		\sum_{e\in X, e\in \gamma} 1_{(f+\gamma)(e)\neq f'(e)} +  1_{(f+\gamma)(e)= f'(e)}
+		\Big).
+	\end{align*}
+	Plugging this in the previous equation, we find that 
+	$$d(f,f')+w(\gamma) = d(f+\gamma, f').$$
+\end{proof}
+
+This result suggests that given some flow $f_i$, we just need to find a negative cycle $\gamma$ in $G_{f_i}$ to construct $f_{i+1}$ as $f_i+\gamma$. The following proposition ensures that this greedy strategy reaches an optimal flow. 
+
+\begin{proposition}
+	For any maximal flow $f$, $G_f$ contains a negative cycle if and only if there exists a maximal flow $f^*$ in $G$ such that $d(f^*, f') < d(f, f')$.
+\end{proposition}
+\begin{proof}
+	Suppose that there is such flow $f^*$. Define the oriented multigraph $M_{f,f^*}=(V,E_M)$ with the same vertex set $V$ as in $G$, and for every $v_1,v_2 \in V$, $E_M$ contains $(f^*(v_1,v_2) - f(v_1,v_2))_+$  copies of the arc $(v_1,v_2)$. For every vertex $v$, its total degree (meaning its outer degree minus its inner degree) is equal to 
+	\begin{align*}
+		\deg v & = \sum_{u\in V} (f^*(v,u) - f(v,u))_+ - \sum_{u\in V} (f^*(u,v) - f(u,v))_+  \\
+		& = \sum_{u\in V} f^*(v,u) - f(v,u) = \sum_{u\in V} f^*(v,u) - \sum_{u\in V}  f(v,u).
+	\end{align*}
+	The last two sums are zero for any inner vertex since $f,f^*$ are flows, and they are equal on the source and sink since the two flows are both maximal and have hence the same value. Thus, $\deg v = 0$ for every vertex $v$.
+	
+	This implies that the multigraph $M_{f,f^*}$ is the union of disjoint simple cycles. $f$ can be transformed into $f^*$ by pushing a mass 1 along all these cycles in any order. Since $d(f^*, f')<d(f,f')$, there must exists one of these simple cycles $\gamma$ with $d(f+\gamma, f') < d(f, f')$. Finally, since we can push a mass in $f$ along $\gamma$, it must appear in $G_f$. Hence $\gamma$ is a cycle of $G_f$ with negative weight.
+\end{proof}
+
+In the next section we describe the corresponding algorithm. Instead of discovering only one cycle, we are allowed to discover a set $\Gamma$ of disjoint negative cycles. 
+
+\subsubsection*{Algorithm}
+\begin{algorithmic}[1]
+	\Function{Minimize transfer load}{$G$, $f$, $\alpha'$}
+	\State Build the graph $G_f$
+	\State $\Gamma \leftarrow$ \Call{Detect Negative Cycles}{$G_f$}
+	\While{$\Gamma \neq \emptyset$}
+	\ForAll{$\gamma \in \Gamma$}
+	\State $f \leftarrow f+\gamma$
+	\EndFor
+	\State Update $G_f$
+	\State $\Gamma \leftarrow$ \Call{Detect Negative Cycles}{$G_f$}
+	\EndWhile
+	\State \Return $f$
+	\EndFunction
+\end{algorithmic}
+
+\subsubsection*{Complexity}
+The distance $d(f,f')$ is bounded by the maximal number of differences in the associated assignment. If these assignment are totally disjoint, this distance is $2\rho_N P$. At every iteration of the While loop, the distance decreases, so there is at most $O(\rho_N P) = O(P)$ iterations.
+
+The detection of negative cycle is done with the Bellman-Ford algorithm, whose complexity should normally be $O(\#E\#V)$. In our case, it amounts to $O(P^2ZN)$. Multiplied by the complexity of the outer loop, it amounts to $O(P^3ZN)$ which is a lot when the number of partitions and nodes starts to be large. To avoid that, we adapt the Bellman-Ford algorithm.
+
+The Bellman-Ford algorithm runs $\#V$ iterations of an outer loop, and an inner loop over $E$. The idea is to compute the shortest paths from a source vertex $v$ to all other vertices. After $k$ iterations of the outer loop, the algorithm has computed all shortest path of length at most $k$. All simple paths have length at most $\#V-1$, so if there is an update in the last iteration of the loop, it means that there is a negative cycle in the graph. The observation that will enable us to improve the complexity is the following:
+
+\begin{proposition}
+	In the graph $G_f$ (and $G$), all simple paths have a length at most $4N$.
+\end{proposition}
+\begin{proof}
+	Since $f$ is a maximal flow, there is no outgoing edge from $\mathbf{s}$ in $G_f$. One can thus check than any simple path of length 4 must contain at least two node of type $\mathbf{n}$. Hence on a path, at most 4 arcs separate two successive nodes of type $\mathbf{n}$. 
+\end{proof}
+
+Thus, in the absence of negative cycles, shortest paths in $G_f$ have length at most $4N$. So we can do only $4N+1$ iterations of the outer loop in the Bellman-Ford algorithm. This makes the complexity of the detection of one set of cycle to be $O(N\#E) = O(N^2 P)$.
+
+With this improvement, the complexity of the whole algorithm is, in the worst case, $O(N^2P^2)$. However, since we detect several cycles at once and we start with a flow that might be close to the previous one, the number of iterations of the outer loop might be smaller in practice.
+
+
+
+\subsubsection*{Metrics}
+We can display the node and zone utilization ratio, by dividing the flow passing through them divided by their outgoing capacity. In particular, we can pinpoint saturated nodes and zones (i.e. used at their full potential).
+
+We can display the distance to the previous assignment, and the number of partition transfers.
+
+
+\bibliography{optimal_layout} 
+\bibliographystyle{ieeetr}
+
+\end{document}
+
+
+
@@ -0,0 +1,11 @@
+
+@article{even1975network,
+	title={Network flow and testing graph connectivity},
+	author={Even, Shimon and Tarjan, R Endre},
+	journal={SIAM journal on computing},
+	volume={4},
+	number={4},
+	pages={507--518},
+	year={1975},
+	publisher={SIAM}
+}
@@ -0,0 +1,709 @@
+\documentclass[]{article}
+
+\usepackage{amsmath,amssymb}
+\usepackage{amsthm}
+
+\usepackage{graphicx,xcolor}
+
+\usepackage{algorithm,algpseudocode,float} 
+
+\renewcommand\thesubsubsection{\Alph{subsubsection})}
+
+\newtheorem{proposition}{Proposition}
+
+%opening
+\title{Optimal partition assignment in Garage}
+\author{Mendes}
+
+\begin{document}
+
+\maketitle
+
+\section{Introduction}
+
+\subsection{Context}
+
+Garage is an open-source distributed storage service blablabla$\dots$
+
+Every object to be stored in the system falls in a partition given by the last $k$ bits of its hash. There are $P=2^k$ partitions. Every partition will be stored on distinct nodes of the system. The goal of the assignment of partitions to nodes is to ensure (nodes and zone) redundancy and to be as efficient as possible.
+
+\subsection{Formal description of the problem}
+
+We are given a set of nodes $\mathbf{N}$ and a set of zones $\mathbf{Z}$. Every node $n$ has a non-negative storage capacity $c_n\ge 0$ and belongs to a zone $z\in \mathbf{Z}$. We are also given a number of partition $P>0$ (typically $P=256$).
+
+We would like to compute an assignment of nodes to partitions. We will impose some redundancy constraints to this assignment, and under these constraints, we want our system to have the largest storage capacity possible. To link storage capacity to partition assignment, we make the following assumption:
+\begin{equation}
+	\tag{H1}
+	\text{\emph{All partitions have the same size $s$.}}
+\end{equation}
+This assumption is justified by the dispersion of the hashing function, when the number of partitions is small relative to the number of stored large objects.
+
+Every node $n$ will store some number $k_n$ of partitions. Hence the partitions stored by $n$ (and hence all partitions by our assumption) have there size bounded by $c_n/k_n$. This remark leads us to define the optimal size that we will want to maximize:
+
+\begin{equation}
+	\label{eq:optimal}
+	\tag{OPT}
+s^* = \min_{n \in N} \frac{c_n}{k_n}.
+\end{equation}
+
+When the capacities of the nodes are updated (this includes adding or removing a node), we want to update the assignment as well. However, transferring the data between nodes has a cost and we would like to limit the number of changes in the assignment. We make the following assumption:
+\begin{equation}
+	\tag{H2}
+	\text{\emph{Updates of capacity happens rarely relatively to object storing.}}
+\end{equation}
+This assumption justifies that when we compute the new assignment, it is worth to optimize the partition size \eqref{eq:optimal} first, and then, among the possible optimal solution, to try to minimize the number of partition transfers.
+
+For now, in the following, we ask the following redundancy constraint:
+
+\textbf{Parametric node and zone redundancy:} Given two integer parameters $1\le \rho_\mathbf{Z} \le \rho_\mathbf{N}$, we ask every partition to be stored on $\rho_\mathbf{N}$ distinct nodes, and these nodes must belong to at least $\rho_\mathbf{Z}$ distinct zones.
+
+
+\textbf{Mode 3-strict:}  every partition needs to be assignated to three nodes belonging to three different zones.
+
+\textbf{Mode 3:} every partition needs to be assignated to three nodes. We try to spread the three nodes over different zones as much as possible.
+
+\textbf{Warning:} This is a working document written incrementally. The last version of the algorithm is the \textbf{parametric assignment} described in the next section.
+
+
+\section{Computation of a parametric assignment}
+\textbf{Attention : }We change notations in this section.
+
+Notations : let $P$ be the number of partitions, $N$ the number of nodes, $Z$ the number of zones. Let $\mathbf{P,N,Z}$ be the label sets of, respectively, partitions, nodes and zones.
+Let $s^*$ be the largest partition size achievable with the redundancy constraints. Let $(c_n)_{n\in \mathbf{N}}$ be the storage capacity of every node. 
+
+In this section, we propose a third specification of the problem. The user inputs two redundancy parameters $1\le \rho_\mathbf{Z} \le \rho_\mathbf{N}$. We compute an assignment $\alpha = (\alpha_p^1, \ldots, \alpha_p^{\rho_\mathbf{N}})_{p\in \mathbf{P}}$ such that every partition $p$ is associated to $\rho_\mathbf{N}$ distinct nodes $\alpha_p^1, \ldots, \alpha_p^{\rho_\mathbf{N}}$ and these nodes belong to at least $\rho_\mathbf{Z}$ distinct zones.
+
+If the layout contained a previous assignment $\alpha'$, we try to minimize the amount of data to transfer during the layout update by making $\alpha$ as close as possible to $\alpha'$.
+
+In the following subsections, we describe the successive steps of the algorithm we propose to compute $\alpha$.
+
+\subsubsection*{Algorithm}
+
+\begin{algorithmic}[1]
+	\Function{Compute Layout}{$\mathbf{N}$, $\mathbf{Z}$, $\mathbf{P}$, $(c_n)_{n\in \mathbf{N}}$, $\rho_\mathbf{N}$, $\rho_\mathbf{Z}$, $\alpha'$}
+	\State $s^* \leftarrow$ \Call{Compute Partition Size}{$\mathbf{N}$, $\mathbf{Z}$, $\mathbf{P}$, $(c_n)_{n\in \mathbf{N}}$, $\rho_\mathbf{N}$, $\rho_\mathbf{Z}$}
+	\State $G \leftarrow G(s^*)$
+	\State $f \leftarrow$ 	\Call{Compute Candidate Assignment}{$G$, $\alpha'$}
+	\State $f^* \leftarrow$ 	\Call{Minimize transfer load}{$G$, $f$, $\alpha'$}
+	\State Build $\alpha^*$ from $f^*$
+	\State \Return $\alpha^*$
+	\EndFunction
+\end{algorithmic}
+
+\subsubsection*{Complexity}
+As we will see in the next sections, the worst case complexity of this algorithm is $O(P^2 N^2)$. The minimization of transfer load is the most expensive step, and it can run with a timeout since it is only an optimization step. Without this step (or with a smart timeout), the worst cas complexity can be $O((PN)^{3/2}\log C)$ where $C$ is the total storage capacity of the cluster. 
+
+\subsection{Determination of the partition size $s^*$}
+
+Again, we will represent an assignment $\alpha$ as a flow in a specific graph $G$. We will not compute the optimal partition size $s^*$ a priori, but we will determine it by dichotomy, as the largest size $s$ such that the maximal flow achievable on $G=G(s)$ has value $\rho_\mathbf{N}P$. We will assume that the capacities are given in a small enough unit (say, Megabytes), and we will determine $s^*$ at the precision of the given unit.
+
+Given some candidate size value $s$, we describe the oriented weighted graph $G=(V,E)$ with vertex set $V$ arc set $E$.
+
+The set of vertices $V$ contains the source $\mathbf{s}$, the sink $\mathbf{t}$, vertices 
+$\mathbf{p^+, p^-}$ for every partition $p$, vertices $\mathbf{x}_{p,z}$ for every partition $p$ and zone $z$, and vertices $\mathbf{n}$ for every node $n$. 
+
+The set of arcs $E$ contains:
+\begin{itemize}
+	\item ($\mathbf{s}$,$\mathbf{p}^+$, $\rho_\mathbf{Z}$) for every partition $p$;
+	\item ($\mathbf{s}$,$\mathbf{p}^-$, $\rho_\mathbf{N}-\rho_\mathbf{Z}$) for every partition $p$;
+	\item ($\mathbf{p}^+$,$\mathbf{x}_{p,z}$, 1) for every partition $p$ and zone $z$;
+	\item ($\mathbf{p}^-$,$\mathbf{x}_{p,z}$, $\rho_\mathbf{N}-\rho_\mathbf{Z}$) for every partition $p$ and zone $z$;
+	\item ($\mathbf{x}_{p,z}$,$\mathbf{n}$, 1) for every partition $p$, zone $z$ and node $n\in z$;
+	\item ($\mathbf{n}$, $\mathbf{t}$, $\lfloor c_n/s \rfloor$) for every node $n$.
+\end{itemize}
+
+In the following complexity calculations, we will use the number of vertices and edges of $G$. Remark from now that $\# V = O(PZ)$ and $\# E = O(PN)$.
+
+\begin{proposition}
+	An assignment $\alpha$ is realizable with partition size $s$ and the redundancy constraints $(\rho_\mathbf{N},\rho_\mathbf{Z})$ if and only if there exists a maximal flow function $f$ in $G$ with total flow $\rho_\mathbf{N}P$, such that the arcs ($\mathbf{x}_{p,z}$,$\mathbf{n}$, 1) used are exactly those for which $p$ is associated to $n$ in $\alpha$.
+\end{proposition}
+\begin{proof}
+	Given such flow $f$, we can reconstruct a candidate $\alpha$. In $f$, the flow passing through $\mathbf{p^+}$ and $\mathbf{p^-}$ is $\rho_\mathbf{N}$, and since the outgoing capacity of every $\mathbf{x}_{p,z}$ is 1, every partition is associated to $\rho_\mathbf{N}$ distinct nodes. The fraction $\rho_\mathbf{Z}$ of the flow passing through every $\mathbf{p^+}$ must be spread over as many distinct zones as every arc outgoing from $\mathbf{p^+}$ has capacity 1. So the reconstructed $\alpha$ verifies the redundancy constraints. For every node $n$, the flow between $\mathbf{n}$ and $\mathbf{t}$ corresponds to the number of partitions associated to $n$. By construction of $f$, this does not exceed $\lfloor c_n/s \rfloor$. We assumed that the partition size is $s$, hence this association does not exceed the storage capacity of the nodes.
+	
+	In the other direction, given an assignment $\alpha$, one can similarly check that the facts that $\alpha$ respects the redundancy constraints, and the storage capacities of the nodes, are necessary condition to construct a maximal flow function $f$.
+\end{proof}
+
+\textbf{Implementation remark:} In the flow algorithm, while exploring the graph, we explore the neighbours of every vertex in a random order to heuristically spread the association between nodes and partitions.
+
+\subsubsection*{Algorithm}
+With this result mind, we can describe the first step of our algorithm. All divisions are supposed to be integer division.
+\begin{algorithmic}[1]
+	\Function{Compute Partition Size}{$\mathbf{N}$, $\mathbf{Z}$, $\mathbf{P}$, $(c_n)_{n\in \mathbf{N}}$, $\rho_\mathbf{N}$, $\rho_\mathbf{Z}$}
+	
+	\State Build the graph $G=G(s=1)$
+	\State $ f \leftarrow$ \Call{Maximal flow}{$G$}
+	\If{$f.\mathrm{total flow} < \rho_\mathbf{N}P$}
+	
+	\State \Return Error: capacities too small or constraints too strong.
+	\EndIf
+	
+	\State $s^- \leftarrow 1$
+	\State $s^+ \leftarrow 1+\frac{1}{\rho_\mathbf{N}}\sum_{n \in \mathbf{N}} c_n$
+	
+	\While{$s^-+1 < s^+$}
+	\State Build the graph $G=G(s=(s^-+s^+)/2)$
+	\State $ f \leftarrow$ \Call{Maximal flow}{$G$}
+	\If{$f.\mathrm{total flow} < \rho_\mathbf{N}P$}
+	\State $s^+ \leftarrow (s^- + s^+)/2$
+	\Else
+	\State $s^- \leftarrow (s^- + s^+)/2$
+	\EndIf
+	\EndWhile
+	
+	\State \Return $s^-$
+	\EndFunction
+\end{algorithmic}
+
+\subsubsection*{Complexity}
+
+To compute the maximal flow, we use Dinic's algorithm. Its complexity on general graphs is $O(\#V^2 \#E)$, but on graphs with edge capacity bounded by a constant, it turns out to be $O(\#E^{3/2})$. The graph $G$ does not fall in this case since the capacities of the arcs incoming to $\mathbf{t}$ are far from bounded. However, the proof of this complexity works readily for graph where we only ask the edges \emph{not} incoming to the sink $\mathbf{t}$ to have their capacities bounded by a constant. One can find the proof of this claim in \cite[Section 2]{even1975network}.
+The dichotomy adds a logarithmic factor $\log (C)$ where $C=\sum_{n \in \mathbf{N}} c_n$ is the total capacity of the cluster. The total complexity of this first function is hence 
+$O(\#E^{3/2}\log C ) = O\big((PN)^{3/2} \log C\big)$.
+
+\subsubsection*{Metrics}
+We can display the discrepancy between the computed $s^*$ and the best size we could hope for a given total capacity, that is $C/\rho_\mathbf{N}$.
+
+\subsection{Computation of a candidate assignment}
+
+Now that we have the optimal partition size $s^*$, to compute a candidate assignment, it would be enough to compute a maximal flow function $f$ on $G(s^*)$. This is what we do if there was no previous assignment $\alpha'$.
+
+If there was some $\alpha'$, we add a step that will heuristically help to obtain a candidate $\alpha$ closer to $\alpha'$. to do so, we fist compute a flow function $\tilde{f}$ that uses only the partition-to-node association appearing in $\alpha'$. Most likely, $\tilde{f}$ will not be a maximal flow of $G(s^*)$. In Dinic's algorithm, we can start from a non maximal flow function and then discover improving paths. This is what we do in starting from $\tilde{f}$. The hope\footnote{This is only a hope, because one can find examples where the construction of $f$ from $\tilde{f}$ produces an assignment $\alpha$ that is not as close as possible to $\alpha'$.} is that the final flow function $f$ will tend to keep the associations appearing in $\tilde{f}$.
+
+More formally, we construct the graph $G_{|\alpha'}$ from $G$ by removing all the arcs $(\mathbf{x}_{p,z},\mathbf{n}, 1)$ where $p$ is not associated to $n$ in $\alpha'$. We compute a maximal flow function $\tilde{f}$ in $G_{|\alpha'}$. $\tilde{f}$ is also a valid (most likely non maximal) flow function in $G$. We compute a maximal flow function $f$ on $G$ by starting Dinic's algorithm on $\tilde{f}$.
+
+\subsubsection*{Algorithm}
+\begin{algorithmic}[1]
+	\Function{Compute Candidate Assignment}{$G$, $\alpha'$}
+	\State Build the graph $G_{|\alpha'}$
+	\State $ \tilde{f} \leftarrow$ \Call{Maximal flow}{$G_{|\alpha'}$}
+	\State $ f \leftarrow$ \Call{Maximal flow from flow}{$G$, $\tilde{f}$}
+	\State \Return $f$
+	\EndFunction
+\end{algorithmic}
+
+\textbf{Remark:} The function ``Maximal flow'' can be just seen as the function ``Maximal flow from flow'' called with the zero flow function as starting flow.
+
+\subsubsection*{Complexity}
+From the consideration of the last section, we have the complexity of the Dinic's algorithm $O(\#E^{3/2}) = O((PN)^{3/2})$.
+
+\subsubsection*{Metrics}
+
+We can display the flow value of $\tilde{f}$, which is an upper bound of the distance between $\alpha$ and $\alpha'$. It might be more a Debug level display than Info.
+
+\subsection{Minimization of the transfer load}
+
+Now that we have a candidate flow function $f$, we want to modify it to make its associated assignment as close as possible to $\alpha'$. Denote by $f'$ the maximal flow associated to $\alpha'$, and let $d(f, f')$ be distance between the associated assignments\footnote{It is the number of arcs of type $(\mathbf{x}_{p,z},\mathbf{n})$ saturated in one flow and not in the other.}. 
+We want to build a sequence $f=f_0, f_1, f_2 \dots$ of maximal flows such that $d(f_i, \alpha')$ decreases as $i$ increases. The distance being a non-negative integer, this sequence of flow functions must be finite. We now explain how to find some improving $f_{i+1}$ from $f_i$.
+
+For any maximal flow $f$ in $G$, we define the oriented weighted graph $G_f=(V, E_f)$ as follows. The vertices of $G_f$ are the same as the vertices of $G$. $E_f$ contains the arc $(v_1,v_2, w)$ between vertices $v_1,v_2\in V$ with weight $w$ if and only if the arc $(v_1,v_2)$ is not saturated in $f$ (i.e. $c(v_1,v_2)-f(v_1,v_2) \ge 1$, we also consider reversed arcs). The weight $w$ is: 
+\begin{itemize}
+	\item  $-1$  if $(v_1,v_2)$ is of type $(\mathbf{x}_{p,z},\mathbf{n})$ or $(\mathbf{x}_{p,z},\mathbf{n})$ and is saturated in only one of the two flows $f,f'$;
+	\item $+1$  if $(v_1,v_2)$ is of type $(\mathbf{x}_{p,z},\mathbf{n})$ or $(\mathbf{x}_{p,z},\mathbf{n})$ and is saturated in either both or none of the two flows $f,f'$;
+	\item $0$ otherwise.
+\end{itemize}
+
+If $\gamma$ is a simple cycle of arcs in $G_f$, we define its weight $w(\gamma)$ as the sum of the weights of its arcs. We can add $+1$ to the value of $f$ on the arcs of $\gamma$, and by construction of $G_f$ and the fact that $\gamma$ is a cycle, the function that we get is still a valid flow function on $G$, it is maximal as it has the same flow value as $f$. We denote this new function $f+\gamma$.
+
+\begin{proposition}
+	Given a maximal flow $f$ and a simple cycle $\gamma$ in $G_f$, we have $d(f+\gamma, f') - d(f,f') = w(\gamma)$.
+\end{proposition}
+\begin{proof}
+	Let $X$ be the set of arcs of type $(\mathbf{x}_{p,z},\mathbf{n})$. Then we can express $d(f,f')$ as
+	\begin{align*}
+		d(f,f') & = \#\{e\in X ~|~ f(e)\neq f'(e)\}
+		= \sum_{e\in X} 1_{f(e)\neq f'(e)} \\
+		& = \frac{1}{2}\big( \#X + \sum_{e\in X} 1_{f(e)\neq f'(e)} - 1_{f(e)= f'(e)} \big).
+	\end{align*}
+	We can express the cycle weight as
+	\begin{align*}
+		w(\gamma) &  = \sum_{e\in X, e\in \gamma} - 1_{f(e)\neq f'(e)} + 1_{f(e)= f'(e)}.
+	\end{align*}
+	Remark that since we passed on unit of flow in $\gamma$ to construct $f+\gamma$, we have for any $e\in X$, $f(e)=f'(e)$ if and only if $(f+\gamma)(e) \neq f'(e)$.
+	Hence
+	\begin{align*}
+		w(\gamma) &  = \frac{1}{2}(w(\gamma) + w(\gamma)) \\
+		&= \frac{1}{2} \Big( 
+		\sum_{e\in X, e\in \gamma} - 1_{f(e)\neq f'(e)} +  1_{f(e)= f'(e)} \\
+		& \qquad +
+		\sum_{e\in X, e\in \gamma} 1_{(f+\gamma)(e)\neq f'(e)} +  1_{(f+\gamma)(e)= f'(e)}
+		\Big).
+	\end{align*}
+	Plugging this in the previous equation, we find that 
+	$$d(f,f')+w(\gamma) = d(f+\gamma, f').$$
+\end{proof}
+
+This result suggests that given some flow $f_i$, we just need to find a negative cycle $\gamma$ in $G_{f_i}$ to construct $f_{i+1}$ as $f_i+\gamma$. The following proposition ensures that this greedy strategy reaches an optimal flow. 
+
+\begin{proposition}
+	For any maximal flow $f$, $G_f$ contains a negative cycle if and only if there exists a maximal flow $f^*$ in $G$ such that $d(f^*, f') < d(f, f')$.
+\end{proposition}
+\begin{proof}
+	Suppose that there is such flow $f^*$. Define the oriented multigraph $M_{f,f^*}=(V,E_M)$ with the same vertex set $V$ as in $G$, and for every $v_1,v_2 \in V$, $E_M$ contains $(f^*(v_1,v_2) - f(v_1,v_2))_+$  copies of the arc $(v_1,v_2)$. For every vertex $v$, its total degree (meaning its outer degree minus its inner degree) is equal to 
+	\begin{align*}
+		\deg v & = \sum_{u\in V} (f^*(v,u) - f(v,u))_+ - \sum_{u\in V} (f^*(u,v) - f(u,v))_+  \\
+		& = \sum_{u\in V} f^*(v,u) - f(v,u) = \sum_{u\in V} f^*(v,u) - \sum_{u\in V}  f(v,u).
+	\end{align*}
+	The last two sums are zero for any inner vertex since $f,f^*$ are flows, and they are equal on the source and sink since the two flows are both maximal and have hence the same value. Thus, $\deg v = 0$ for every vertex $v$.
+	
+	This implies that the multigraph $M_{f,f^*}$ is the union of disjoint simple cycles. $f$ can be transformed into $f^*$ by pushing a mass 1 along all these cycles in any order. Since $d(f^*, f')<d(f,f')$, there must exists one of these simple cycles $\gamma$ with $d(f+\gamma, f') < d(f, f')$. Finally, since we can push a mass in $f$ along $\gamma$, it must appear in $G_f$. Hence $\gamma$ is a cycle of $G_f$ with negative weight.
+\end{proof}
+
+In the next section we describe the corresponding algorithm. Instead of discovering only one cycle, we are allowed to discover a set $\Gamma$ of disjoint negative cycles. 
+
+\subsubsection*{Algorithm}
+\begin{algorithmic}[1]
+	\Function{Minimize transfer load}{$G$, $f$, $\alpha'$}
+	\State Build the graph $G_f$
+	\State $\Gamma \leftarrow$ \Call{Detect Negative Cycles}{$G_f$}
+	\While{$\Gamma \neq \emptyset$}
+	\ForAll{$\gamma \in \Gamma$}
+	\State $f \leftarrow f+\gamma$
+	\EndFor
+	\State Update $G_f$
+	\State $\Gamma \leftarrow$ \Call{Detect Negative Cycles}{$G_f$}
+	\EndWhile
+	\State \Return $f$
+	\EndFunction
+\end{algorithmic}
+
+\subsubsection*{Complexity}
+The distance $d(f,f')$ is bounded by the maximal number of differences in the associated assignment. If these assignment are totally disjoint, this distance is $2\rho_N P$. At every iteration of the While loop, the distance decreases, so there is at most $O(\rho_N P) = O(P)$ iterations.
+
+The detection of negative cycle is done with the Bellman-Ford algorithm, whose complexity should normally be $O(\#E\#V)$. In our case, it amounts to $O(P^2ZN)$. Multiplied by the complexity of the outer loop, it amounts to $O(P^3ZN)$ which is a lot when the number of partitions and nodes starts to be large. To avoid that, we adapt the Bellman-Ford algorithm.
+
+The Bellman-Ford algorithm runs $\#V$ iterations of an outer loop, and an inner loop over $E$. The idea is to compute the shortest paths from a source vertex $v$ to all other vertices. After $k$ iterations of the outer loop, the algorithm has computed all shortest path of length at most $k$. All simple paths have length at most $\#V-1$, so if there is an update in the last iteration of the loop, it means that there is a negative cycle in the graph. The observation that will enable us to improve the complexity is the following:
+
+\begin{proposition}
+	In the graph $G_f$ (and $G$), all simple paths have a length at most $4N$.
+\end{proposition}
+\begin{proof}
+	Since $f$ is a maximal flow, there is no outgoing edge from $\mathbf{s}$ in $G_f$. One can thus check than any simple path of length 4 must contain at least two node of type $\mathbf{n}$. Hence on a path, at most 4 arcs separate two successive nodes of type $\mathbf{n}$. 
+\end{proof}
+
+Thus, in the absence of negative cycles, shortest paths in $G_f$ have length at most $4N$. So we can do only $4N+1$ iterations of the outer loop in Bellman-Ford algorithm. This makes the complexity of the detection of one set of cycle to be $O(N\#E) = O(N^2 P)$.
+
+With this improvement, the complexity of the whole algorithm is, in the worst case, $O(N^2P^2)$. However, since we detect several cycles at once and we start with a flow that might be close to the previous one, the number of iterations of the outer loop might be smaller in practice.
+
+
+
+\subsubsection*{Metrics}
+We can display the node and zone utilization ratio, by dividing the flow passing through them divided by their outgoing capacity. In particular, we can pinpoint saturated nodes and zones (i.e. used at their full potential).
+
+We can display the distance to the previous assignment, and the number of partition transfers.
+
+
+
+
+
+\section{Properties of an optimal 3-strict assignment}
+
+\subsection{Optimal assignment}
+\label{sec:opt_assign}
+
+For every zone $z\in Z$, define the zone capacity $c_z = \sum_{v, z_v=z} c_v$ and define $C = \sum_v c_v = \sum_z c_z$.
+
+One can check that the best we could be doing to maximize $s^*$ would be to use the nodes proportionally to their capacity. This would yield $s^*=C/(3N)$. This is not possible because of (i) redundancy constraints and (ii) integer rounding but it gives and upper bound. 
+
+\subsubsection*{Optimal utilization}
+
+We call an \emph{utilization} a collection of non-negative integers $(n_v)_{v\in V}$ such that $\sum_v n_v = 3N$ and for every zone $z$, $\sum_{v\in z} n_v \le N$. We call such utilization \emph{optimal} if it maximizes $s^*$.
+
+We start by computing a node sub-utilization $(\hat{n}_v)_{v\in V}$ such that for every zone $z$, $\sum_{v\in z} \hat{n}_v \le N$ and we show that there is an optimal utilization respecting the constraints and such that $\hat{n}_v \le n_v$ for every node.
+
+Assume that there is a zone $z_0$ such that $c_{z_0}/C \ge 1/3$. Then for any $v\in z_0$, we define 
+$$\hat{n}_v = \left\lfloor\frac{c_v}{c_{z_0}}N\right\rfloor.$$ 
+This choice ensures for any such $v$ that
+$$
+\frac{c_v}{\hat{n}_v} \ge \frac{c_{z_0}}{N} \ge \frac{C}{3N}
+$$
+which is the universal upper bound on $s^*$. Hence any optimal utilization $(n_v)$ can be modified to another optimal utilization such that $n_v\ge \hat{n}_v$
+
+Because $z_0$ cannot store more than $N$ partition occurrences, in any assignment, at least $2N$ partitions must be assignated to the zones $Z\setminus\{z_0\}$. Let $C_0 = C-c_{z_0}$. Suppose that there exists a zone $z_1\neq z_0$ such that $c_{z_1}/C_0 \ge 1/2$. Then, with the same argument as for $z_0$, we can define
+$$\hat{n}_v = \left\lfloor\frac{c_v}{c_{z_1}}N\right\rfloor$$
+for every $v\in z_1$.
+
+Now we can assign the remaining partitions. Let $(\hat{N}, \hat{C})$ to be
+\begin{itemize}
+	\item $(3N,C)$ if we did not find any $z_0$;
+	\item $(2N,C-c_{z_0})$ if there was a $z_0$ but no $z_1$;
+	\item $(N,C-c_{z_0}-c_{z_1})$ if there was a $z_0$ and a $z_1$.
+\end{itemize}
+Then at least $\hat{N}$ partitions must be spread among the remaining zones. Hence $s^*$ is upper bounded by $\hat{C}/\hat{N}$ and without loss of generality, we can define, for every node that is not in $z_0$ nor $z_1$, 
+$$\hat{n}_v = \left\lfloor\frac{c_v}{\hat{C}}\hat{N}\right\rfloor.$$ 
+
+We constructed a sub-utilization $\hat{n}_v$. Now notice that $3N-\sum_v \hat{n}_v \le \# V$ where $\# V$ denotes the number of nodes. We can iteratively pick a node $v^*$ such that 
+\begin{itemize}
+	\item $\sum_{v\in z_{v^*}} \hat{n}_v < N$ where $z_{v^*}$ is the zone of $v^*$;
+	\item $v^*$ maximizes the quantity $c_v/(\hat{n}_v+1)$ among the vertices satisfying the first condition (i.e. not in a saturated zone).
+\end{itemize}
+We iterate these instructions until $\sum_v \hat{n}_v= 3N$, and at this stage we define $(n_v) = (\hat{n}_v)$. It is easy to prove by induction that at every step, there is an optimal utilization that is pointwise larger than $\hat{n}_v$, and in particular, that $(n_v)$ is optimal.
+
+\subsubsection*{Existence of an optimal assignment}
+
+As for now, the \emph{optimal utilization} that we obtained is just a vector of numbers and it is not clear that it can be realized as the utilization of some concrete assignment. Here is a way to get a concrete assignment.
+
+Define $3N$ tokens $t_1,\ldots, t_{3N}\in V$ as follows:
+\begin{itemize}
+	\item Enumerate the zones $z$ of $Z$ in any order;
+	\item enumerate the nodes $v$ of $z$ in any order;
+	\item repeat $n_v$ times the token $v$.
+\end{itemize} 
+Then for $1\le i \le N$, define the triplet $T_i$ to be 
+$(t_i, t_{i+N}, t_{i+2N})$. Since the same nodes of a zone appear contiguously, the three nodes of a triplet must belong to three distinct zones.
+
+However simple, this solution to go from an utilization to an assignment has the drawback of not spreading the triplets: a node will tend to be associated to the same two other nodes for many partitions. Hence, during data transfer, it will tend to use only two link, instead of spreading the bandwidth use over many other links to other nodes. To achieve this goal, we will reframe the search of an assignment as a flow problem. and in the flow algorithm, we will introduce randomness in the order of exploration. This will be sufficient to obtain a good dispersion of the triplets.
+
+\begin{figure}
+	\centering
+	\includegraphics[width=0.9\linewidth]{figures/naive}
+	\caption{On the left, the creation of a concrete assignment with the naive approach of repeating tokens. On the right, the zones containing the nodes.}
+\end{figure}
+
+\subsubsection*{Assignment as a maximum flow problem}
+
+We describe the flow problem via its graph $(X,E)$ where $X$ is a set of vertices, and $E$ are directed weighted edges between the vertices. For every zone $z$, define $n_z=\sum_{v\in z} n_v$.
+
+The set of vertices $X$ contains the source $\mathbf{s}$ and the sink $\mathbf{t}$; a vertex $\mathbf{x}_z$ for every zone $z\in Z$, and a vertex $\mathbf{y}_i$ for every partition index $1\le i\le N$. 
+
+The set of edges $E$ contains
+\begin{itemize}
+	\item the edge $(\mathbf{s}, \mathbf{x}_z, n_z)$ for every zone $z\in Z$;
+	\item the edge $(\mathbf{x}_z, \mathbf{y}_i, 1)$ for every zone $z\in Z$ and partition $1\le i\le N$;
+	\item the edge $(\mathbf{y}_i, \mathbf{t}, 3)$ for every partition $1\le i\le N$.
+\end{itemize}
+
+\begin{figure}[b]
+	\centering
+	\includegraphics[width=0.6\linewidth]{figures/flow}
+	\caption{Flow problem to compute and optimal assignment.}
+\end{figure}
+
+We first show the equivalence between this problem and and the construction of an assignment. Given some optimal assignment $(n_v)$, define the flow $f:E\to \mathbb{N}$ that saturates every edge from $\mathbf{s}$ or to $\mathbf{t}$, takes value $1$ on the edge between $\mathbf{x}_z$ and $\mathbf{y}_i$ if partition $i$ is stored in some node of the zone $z$, and $0$ otherwise. One can easily check that $f$ thus defined is indeed a flow and is maximum.
+
+Reciprocally, by the existence of maximum flows constructed from optimal assignments, any maximum flow must saturate the edges linked to the source or the sink. It can only take value 0 or 1 on the other edge, and every partition vertex is associated to exactly three distinct zone vertices. Every zone is associated to exactly $n_z$ partitions.
+
+A maximum flow can be constructed using, for instance, Dinic's algorithm. This algorithm works by discovering augmenting path to iteratively increase the flow. During the exploration of the graph to find augmenting path, we can shuffle the order of enumeration of the neighbours to spread the associations between zones and partitions.
+
+Once we have such association, we can randomly distribute the $n_z$ edges picked for every zone $z$ to its nodes $v\in z$ such that every such $v$ gets $n_z$ edges. This defines an optimal assignment of partitions to nodes.
+
+
+\subsection{Minimal transfer}
+
+Assume that there was a previous assignment $(T'_i)_{1\le i\le N}$ corresponding to utilizations $(n'_v)_{v\in V}$. We would like the new computed assignment $(T_i)_{1\le i\le N}$ from some $(n_v)_{v\in V}$ to minimize the number of partitions that need to be transferred. We can imagine two different objectives corresponding to different hypotheses: 
+\begin{equation}
+	\tag{H3A}
+	\label{hyp:A}
+	\text{\emph{Transfers between different zones cost much more than inside a zone.}}
+\end{equation}
+\begin{equation}
+	\tag{H3B}
+	\label{hyp:B}
+	\text{\emph{Changing zone is not the largest cost when transferring a partition.}}
+\end{equation}
+
+In case $A$, our goal will be to minimize the number of changes of zone in the assignment of partitions to zone. More formally, we will maximize the quantity
+$$
+Q_Z := 
+\sum_{1\le i\le N} 
+\#\{z\in Z ~|~ z\cap T_i \neq \emptyset, z\cap T'_i \neq \emptyset \}
+.$$
+
+In case $B$, our goal will be to minimize the number of changes of nodes in the assignment of partitions to nodes. We will maximize the quantity
+$$
+Q_V :=
+\sum_{1\le i\le N} \#(T_i \cap T'_i).
+$$
+
+It is tempting to hope that there is a way to maximize both quantity, that having the least discrepancy in terms of nodes will lead to the least discrepancy in terms of zones. But this is actually wrong! We propose the following counter-example to convince the reader:
+
+We consider eight nodes $a, a', b, c, d, d', e, e'$ belonging to five different zones $\{a,a'\}, \{b\}, \{c\}, \{d,d'\}, \{e, e'\}$. We take three partitions ($N=3$), that are originally assigned with some utilization $(n'_v)_{v\in V}$ as follows:
+$$
+T'_1=(a,b,c) \qquad
+T'_2=(a',b,d) \qquad 
+T'_3=(b,c,e).
+$$
+This assignment, with updated utilizations $(n_v)_{v\in V}$ minimizes the number of zone changes:
+$$
+T_1=(d,b,c) \qquad
+T_2=(a,b,d) \qquad 
+T_3=(b,c,e').
+$$
+This one, with the same utilization, minimizes the number of node changes:
+$$
+T_1=(a,b,c) \qquad
+T_2=(e',b,d) \qquad 
+T_3=(b,c,d').
+$$
+One can check that in this case, it is impossible to minimize both the number of zone and node changes.
+
+Because of the redundancy constraint, we cannot use a greedy algorithm to just replace nodes in the triplets to try to get the new utilization rate: this could lead to blocking situation where there is still a hole to fill in a triplet but no available node satisfies the zone separation constraint. To circumvent this issue, we propose an algorithm based on finding cycles in a graph encoding of the assignment. As in section \ref{sec:opt_assign}, we can explore the neighbours in a random order in the graph algorithms, to spread the triplets distribution.
+
+
+\subsubsection{Minimizing the zone discrepancy}
+
+
+First, notice that, given an assignment of partitions to \emph{zones}, it is easy to deduce an assignment to \emph{nodes} that minimizes the number of transfers for this zone assignment: For every zone $z$ and every node $v\in z$, pick in any way a set $P_v$ of partitions that where assigned to $v$ in $T'$, to $z_v$ in $T$, with the cardinality of $P_v$ smaller than $n_v$. Once all these sets are chosen, complement the assignment to reach the right utilization for every node. If $\#P_v > n_v$, it means that all the partitions that could stay in $v$ (i.e. that were already in $v$ and are still assigned to its zone) do stay in $v$. If $\#P_v = n_v$, then $n_v$ partitions stay in $v$, which is the number of partitions that need to be in $v$ in the end. In both cases, we could not hope for better given the partition to zone assignment.
+
+Our goal now is to find a assignment of partitions to zones that minimizes the number of zone transfers. To do so we are going to represent an assignment as a graph.
+
+Let $G_T=(X,E_T)$ be the directed weighted graph with vertices $(\mathbf{x}_i)_{1\le i\le N}$ and $(\mathbf{y}_z)_{z\in Z}$. For any $1\le i\le N$ and $z\in Z$, $E_T$ contains  the arc:
+\begin{itemize}
+	\item $(\mathbf{x}_i, \mathbf{y}_z, +1)$, if $z$ appears in $T_i'$ and $T_i$;
+	\item $(\mathbf{x}_i, \mathbf{y}_z, -1)$, if $z$ appears in $T_i$ but not in $T'_i$;
+	\item $(\mathbf{y}_z, \mathbf{x}_i, -1)$, if $z$ appears in $T'_i$ but not in $T_i$;
+	\item $(\mathbf{y}_z, \mathbf{x}_i, +1)$, if $z$ does not appear in $T'_i$ nor in $T_i$.
+\end{itemize}
+In other words, the orientation of the arc encodes whether partition $i$ is stored in zone $z$ in the assignment $T$ and the weight $\pm 1$ encodes whether this corresponds to what happens in the assignment $T'$.
+
+\begin{figure}[t]
+	\centering
+	\begin{minipage}{.40\linewidth}
+		\centering
+		\includegraphics[width=.8\linewidth]{figures/mini_zone}
+	\end{minipage}
+	\begin{minipage}{.55\linewidth}
+		\centering
+		\includegraphics[width=.8\linewidth]{figures/mini_node}
+	\end{minipage}
+	\caption{On the left: the graph $G_T$ encoding an assignment to minimize the zone discrepancy. On the right: the graph $G_T$ encoding an assignment to minimize the node discrepancy.}
+\end{figure}
+
+
+Notice that at every partition, there are three outgoing arcs, and at every zone, there are $n_z$ incoming arcs. Moreover, if $w(e)$ is the weight of an arc $e$, define the weight of $G_T$ by
+\begin{align*}
+w(G_T) := \sum_{e\in E} w(e) &= \#Z \times N - 4 \sum_{1\le i\le N} \#\{z\in Z ~|~ z\cap T_i = \emptyset, z\cap T'_i \neq \emptyset\}	\\
+&=\#Z \times N - 4 \sum_{1\le i\le N} 3- \#\{z\in Z ~|~ z\cap T_i \neq \emptyset, z\cap T'_i \neq \emptyset\} \\
+&= (\#Z-12)N + 4 Q_Z.
+\end{align*}
+Hence maximizing $Q_Z$ is equivalent to maximizing $w(G_T)$.
+
+Assume that their exist some assignment $T^*$ with the same utilization $(n_v)_{v\in V}$. Define $G_{T^*}$ similarly and consider the set $E_\mathrm{Diff} = E_T \setminus E_{T^*}$ of arcs that appear only in $G_T$. Since all vertices have the same number of incoming arcs in $G_T$ and $G_{T^*}$, the vertices of the graph $(X, E_\mathrm{Diff})$ must all have the same number number of incoming and outgoing arrows. So $E_\mathrm{Diff}$ can be expressed as a union of disjoint cycles. Moreover, the edges of $E_\mathrm{Diff}$ must appear in $E_{T^*}$ with reversed orientation and opposite weight. Hence, we have
+$$
+ w(G_T) - w(G_{T^*}) = 2 \sum_{e\in E_\mathrm{Diff}} w(e). 
+$$
+Hence, if $T$ is not optimal, there exists some $T^*$ with $w(G_T) < w(G_{T^*})$, and by the considerations above, there must exist a cycle in $E_\mathrm{Diff}$, and hence in $G_T$, with negative weight. If we reverse the edges and weights along this cycle, we obtain some graph. Since we did not change the incoming degree of any vertex, this is the graph encoding of some valid assignment $T^+$ such that $w(G_{T^+}) > w(G_T)$. We can iterate this operation until there is no other assignment $T^*$ with larger weight, that is until we obtain an optimal assignment.
+
+
+
+\subsubsection{Minimizing the node discrepancy}
+
+We will follow an approach similar to the one where we minimize the zone discrepancy. Here we will directly obtain a node assignment from a graph encoding.
+
+Let $G_T=(X,E_T)$ be the directed weighted graph with vertices $(\mathbf{x}_i)_{1\le i\le N}$, $(\mathbf{y}_{z,i})_{z\in Z, 1\le i\le N}$ and $(\mathbf{u}_v)_{v\in V}$. For any $1\le i\le N$ and $z\in Z$, $E_T$ contains  the arc:
+\begin{itemize}
+	\item $(\mathbf{x}_i, \mathbf{y}_{z,i}, 0)$, if $z$ appears in $T_i$;
+	\item $(\mathbf{y}_{z,i}, \mathbf{x}_i, 0)$, if $z$ does not appear in $T_i$.
+\end{itemize}
+For any $1\le i\le N$ and $v\in V$, $E_T$ contains  the arc:
+\begin{itemize}
+	\item $(\mathbf{y}_{z_v,i}, \mathbf{u}_v, +1)$, if $v$ appears in $T_i'$ and $T_i$;
+	\item $(\mathbf{y}_{z_v,i}, \mathbf{u}_v, -1)$, if $v$ appears in $T_i$ but not in $T'_i$;
+	\item $(\mathbf{u}_v, \mathbf{y}_{z_v,i}, -1)$, if $v$ appears in $T'_i$ but not in $T_i$;
+	\item $(\mathbf{u}_v, \mathbf{y}_{z_v,i}, +1)$, if $v$ does not appear in $T'_i$ nor in $T_i$.
+\end{itemize}
+Every vertex $\mathbb{x}_i$ has outgoing degree 3, every vertex $\mathbf{y}_{z,v}$ has outgoing degree 1, and every vertex $\mathbf{u}_v$ has incoming degree $n_v$. 
+Remark that any graph respecting these degree constraints is the encoding of a valid assignment with utilizations $(n_v)_{v\in V}$, in particular no partition is stored in two nodes of the same zone.
+
+We define $w(G_T)$ similarly:
+\begin{align*}
+	w(G_T) := \sum_{e\in E_T} w(e) &= \#V \times N - 4\sum_{1\le i\le N} 3-\#(T_i\cap T'_i) \\
+	&= (\#V-12)N + 4Q_V.
+\end{align*}
+
+Exactly like in the previous section, the existence of an assignment with larger weight implies the existence of a negatively weighted cycle in $G_T$. Reversing this cycle gives us the encoding of a valid assignment with a larger weight. Iterating this operation yields an optimal assignment.
+
+
+\subsubsection{Linear combination of both criteria}
+
+In the graph $G_T$ defined in the previous section, instead of having weights $0$ and $\pm 1$, we could be having weights $\pm\alpha$ between $\mathbf{x}$ and  $\mathbf{y}$ vertices, and weights $\pm\beta$ between $\mathbf{y}$ and  $\mathbf{u}$ vertices, for some $\alpha,\beta>0$ (we have positive weight if the assignment corresponds to $T'$ and negative otherwise). Then
+\begin{align*}
+	w(G_T) &= \sum_{e\in E_T} w(e) = 
+	\alpha \big( (\#Z-12)N + 4 Q_Z\big) + 
+	\beta \big( (\#V-12)N + 4 Q_V\big) \\
+	&= \mathrm{const}+ 4(\alpha Q_Z + \beta Q_V).
+\end{align*}
+So maximizing the weight of such graph encoding would be equivalent to maximizing a linear combination of $Q_Z$ and $Q_V$.
+
+
+\subsection{Algorithm}
+We give a high level description of the algorithm to compute an optimal 3-strict assignment. The operations appearing at lines 1,2,4 are respectively described by Algorithms \ref{alg:util},\ref{alg:opt} and \ref{alg:mini}.
+
+
+
+\begin{algorithm}[H]
+	\caption{Optimal 3-strict assignment}
+	\label{alg:total}
+	\begin{algorithmic}[1]
+	\Function{Optimal 3-strict assignment}{$N$, $(c_v)_{v\in V}$, $T'$}
+		\State $(n_v)_{v\in V} \leftarrow$ \Call{Compute optimal utilization}{$N$, $(c_v)_{v\in V}$}
+		\State $(T_i)_{1\le i\le N} \leftarrow$ \Call{Compute candidate assignment}{$N$, $(n_v)_{v\in V}$}
+		\If {there was a previous assignment $T'$}
+			\State $T \leftarrow$ \Call{Minimization of transfers}{$(T_i)_{1\le i\le N}$, $(T'_i)_{1\le i\le N}$}
+		\EndIf
+		\State \Return $T$.
+	\EndFunction
+	\end{algorithmic}
+\end{algorithm}
+
+We give some considerations of worst case complexity for these algorithms. In the following, we assume $N>\#V>\#Z$. The complexity of Algorithm \ref{alg:total} is $O(N^3\# Z)$ if we assume \eqref{hyp:A} and $O(N^3 \#Z \#V)$ if we assume \eqref{hyp:B}.
+
+Algorithm \ref{alg:util} can be implemented with complexity $O(\#V^2)$. The complexity of the function call at line \ref{lin:subutil} is $O(\#V)$. The difference between the sum of the subutilizations and $3N$ is at most the sum of the rounding errors when computing the $\hat{n}_v$. Hence it is bounded by $\#V$ and the loop at line \ref{lin:loopsub} is iterated at most $\#V$ times. Finding the minimizing $v$ at line \ref{lin:findmin} takes $O(\#V)$ operations (naively, we could also use a heap).
+
+Algorithm \ref{alg:opt} can be implemented with complexity $O(N^3\times \#Z)$. The flow graph has $O(N+\#Z)$ vertices and $O(N\times \#Z)$ edges. Dinic's algorithm has complexity $O(\#\mathrm{Vertices}^2\#\mathrm{Edges})$ hence in our case it is $O(N^3\times \#Z)$.
+
+Algorithm \ref{alg:mini} can be implemented with complexity $O(N^3\# Z)$ under \eqref{hyp:A} and $O(N^3 \#Z \#V)$ under \eqref{hyp:B}.
+The graph $G_T$ has $O(N)$ vertices and $O(N\times \#Z)$ edges under assumption \eqref{hyp:A} and respectively $O(N\times \#Z)$ vertices and $O(N\times \#V)$ edges under assumption \eqref{hyp:B}. The loop at line \ref{lin:repeat} is iterated at most $N$ times since the distance between $T$ and $T'$ decreases at every iteration. Bellman-Ford algorithm has complexity $O(\#\mathrm{Vertices}\#\mathrm{Edges})$, which in our case amounts to $O(N^2\# Z)$ under \eqref{hyp:A} and $O(N^2 \#Z \#V)$ under \eqref{hyp:B}.
+ 
+\begin{algorithm}
+	\caption{Computation of the optimal utilization}
+	\label{alg:util}
+	\begin{algorithmic}[1]
+\Function{Compute optimal utilization}{$N$, $(c_v)_{v\in V}$}
+	\State $(\hat{n}_v)_{v\in V} \leftarrow $ \Call{Compute subutilization}{$N$, $(c_v)_{v\in V}$} \label{lin:subutil}
+	\While{$\sum_{v\in V} \hat{n}_v < 3N$} \label{lin:loopsub}
+	\State Pick $v\in V$ minimizing $\frac{c_v}{\hat{n}_v+1}$ and such that
+		$\sum_{v'\in z_v} \hat{n}_{v'} < N$ \label{lin:findmin}
+		\State $\hat{n}_v \leftarrow \hat{n}_v+1$
+	\EndWhile
+	\State \Return $(\hat{n}_v)_{v\in V}$
+\EndFunction
+\State
+
+\Function{Compute subutilization}{$N$, $(c_v)_{v\in V}$}
+		\State $R \leftarrow 3$
+\For{$v\in V$}
+\State $\hat{n}_v \leftarrow \mathrm{unset}$
+\EndFor
+\For{$z\in Z$}
+\State $c_z \leftarrow \sum_{v\in z} c_v$
+\EndFor
+\State $C \leftarrow \sum_{z\in Z} c_z$
+\While{$\exists z \in Z$ such that $R\times c_{z} > C$}
+\For{$v\in z$}
+\State $\hat{n}_v \leftarrow \left\lfloor \frac{c_v}{c_z} N \right\rfloor$
+\EndFor
+\State $C \leftarrow C-c_z$
+\State $R\leftarrow R-1$
+\EndWhile
+\For{$v\in V$}
+\If{$\hat{n}_v = \mathrm{unset}$}
+\State $\hat{n}_v \leftarrow \left\lfloor \frac{Rc_v}{C} N \right\rfloor$
+\EndIf
+\EndFor
+\State \Return $(\hat{n}_v)_{v\in V}$
+\EndFunction
+	\end{algorithmic}
+\end{algorithm}
+
+\begin{algorithm}
+	\caption{Computation of a candidate assignment}
+	\label{alg:opt}
+	\begin{algorithmic}[1]
+		\Function{Compute candidate assignment}{$N$, $(n_v)_{v\in V}$}
+		\State Compute the flow graph $G$
+		\State Compute the maximal flow $f$ using Dinic's algorithm with randomized neighbours enumeration
+		\State Construct the assignment $(T_i)_{1\le i\le N}$ from $f$
+		\State \Return $(T_i)_{1\le i\le N}$
+		\EndFunction
+	\end{algorithmic}
+\end{algorithm}
+
+
+\begin{algorithm}
+	\caption{Minimization of the number of transfers}
+	\label{alg:mini}
+	\begin{algorithmic}[1]
+		\Function{Minimization of transfers}{$(T_i)_{1\le i\le N}$, $(T'_i)_{1\le i\le N}$}
+		\State Construct the graph encoding $G_T$
+		\Repeat \label{lin:repeat}
+		\State Find a negative cycle $\gamma$ using Bellman-Ford algorithm on $G_T$
+		\State Reverse the orientations and weights of edges in $\gamma$
+		\Until{no negative cycle is found}
+		\State Update $(T_i)_{1\le i\le N}$ from $G_T$
+		\State \Return $(T_i)_{1\le i\le N}$
+		\EndFunction
+	\end{algorithmic}
+\end{algorithm}
+
+\newpage
+
+\section{Computation of a 3-non-strict assignment}
+
+\subsection{Choices of optimality}
+
+In this mode, we primarily want to store every partition on three nodes, and only secondarily try to spread the nodes among different zone. So we make the choice of not taking the zone repartition in the criterion of optimality.
+
+We try to maximize $s^*$ defined in \eqref{eq:optimal}. So we can compute the optimal utilizations $(n_v)_{v\in V}$ with the only constraint that $n_v \le N$ for every node $v$. As in the previous section, we start with a sub-utilization proportional to $c_v$ (and capped at $N$), and we iteratively increase the $\hat{n}_v$ that is less than $N$ and maximizes the quantity $c_v/(\hat{n}_v+1)$, until the total sum is $3N$.
+
+\subsection{Computation of a candidate assignment}
+
+To compute a candidate assignment (that does not optimize zone spreading nor distance to a previous assignment yet), we can use the following flow problem.
+
+Define the oriented weighted graph $(X,E)$. The set of vertices $X$ contains the source $\mathbf{s}$, the sink $\mathbf{t}$, vertices 
+$\mathbf{x}_p, \mathbf{u}^+_p, \mathbf{u}^-_p$ for every partition $p$, vertices $\mathbf{y}_{p,z}$ for every partition $p$ and zone $z$, and vertices $\mathbf{z}_v$ for every node $v$. 
+
+The set of edges is composed of the following arcs:
+\begin{itemize}
+	\item ($\mathbf{s}$,$\mathbf{x}_p$, 3) for every partition $p$;
+	\item ($\mathbf{x}_p$,$\mathbf{u}^+_p$, 3) for every partition $p$;
+	\item ($\mathbf{x}_p$,$\mathbf{u}^-_p$, 2) for every partition $p$;
+	\item ($\mathbf{u}^+_p$,$\mathbf{y}_{p,z}$, 1) for every partition $p$ and zone $z$;
+	\item ($\mathbf{u}^-_p$,$\mathbf{y}_{p,z}$, 2) for every partition $p$ and zone $z$;
+	\item ($\mathbf{y}_{p,z}$,$\mathbf{z}_v$, 1) for every partition $p$, zone $z$ and node $v\in z$;
+	\item ($\mathbf{z}_v$, $\mathbf{t}$, $n_v$) for every node $v$;
+\end{itemize}
+
+One can check that any maximal flow in this graph corresponds to an assignment of partitions to nodes. In such a flow, all the arcs from $\mathbf{s}$ and to $\mathbf{t}$ are saturated. The arc from $\mathbf{y}_{p,z}$ to $\mathbf{z}_v$ is saturated if and only if $p$ is associated to~$v$. 
+Finally the flow from $\mathbf{x}_p$ to $\mathbf{y}_{p,z}$ can go either through $\mathbf{u}^+_p$ or $\mathbf{u}^-_p$.
+
+
+
+\subsection{Maximal spread and minimal transfers}
+Notice that if the arc $\mathbf{u}_p^+\mathbf{y}_{p,z}$ is not saturated but there is some flow in $\mathbf{u}_p^-\mathbf{y}_{p,z}$, then it is possible to transfer a unit of flow from the path  $\mathbf{x}_p\mathbf{u}_p^-\mathbf{y}_{p,z}$ to the path $\mathbf{x}_p\mathbf{u}_p^+\mathbf{y}_{p,z}$. So we can always find an equivalent maximal flow $f^*$ that uses the path through $\mathbf{u}_p^-$ only if the path through $\mathbf{u}_p^+$ is saturated.
+
+We will use this fact to consider the amount of flow going through the vertices $\mathbf{u}^+$ as a measure of how well the partitions are spread over nodes belonging to different zones. If the partition $p$ is associated to 3 different zones, then a flow of 3 will cross $\mathbf{u}_p^+$ in $f^*$ (i.e. a flow of 0 will cross $\mathbf{u}_p^+$). If $p$ is associated to two zones, a flow of $2$ will cross $\mathbf{u}_p^+$. If $p$ is associated to a single zone, a flow of $1$ will cross $\mathbf{u}_p^+$.
+
+Let $N_1, N_2, N_3$ be the number of partitions associated to respectively 1,2 and 3 distinct zones. We will optimize a linear combination of these variables using the discovery of positively weighted circuits in a graph.
+
+At the same step, we will also optimize the distance to a previous assignment $T'$. Let $\alpha> \beta> \gamma \ge 0$ be three parameters.
+
+Given the flow $f$, let $G_f=(X',E_f)$ be the multi-graph where $X' = X\setminus\{\mathbf{s},\mathbf{t}\}$. The set $E_f$ is composed of the arcs:
+\begin{itemize}
+\item As many arcs from $(\mathbf{x}_p, \mathbf{u}^+_p,\alpha), (\mathbf{x}_p, \mathbf{u}^+_p,\beta), (\mathbf{x}_p, \mathbf{u}^+_p,\gamma)$ (selected in this order) as there is flow crossing $\mathbf{u}^+_p$ in $f$;
+\item As many arcs from $(\mathbf{u}^+_p, \mathbf{x}_p,-\gamma), (\mathbf{u}^+_p, \mathbf{x}_p,-\beta), (\mathbf{u}^+_p, \mathbf{x}_p,-\alpha)$ (selected in this order) as there is flow crossing $\mathbf{u}^-_p$ in $f$;
+\item  As many copies of $(\mathbf{x}_p, \mathbf{u}^-_p,0)$ as there is flow through $\mathbf{u}^-_p$;
+\item  As many copies of $(\mathbf{u}^-_p,\mathbf{x}_p,0)$ so that the number of arcs between these two vertices is 2;
+\item $(\mathbf{u}^+_p,\mathbf{y}_{p,z}, 0)$ if the flow between these vertices is 1, and the opposite arc otherwise;
+\item as many copies of $(\mathbf{u}^-_p,\mathbf{y}_{p,z}, 0)$ as the flow between these vertices, and as many copies of the opposite arc as 2~$-$~the flow;
+\item $(\mathbf{y}_{p,z},\mathbf{z}_v, \pm1)$ if it is saturated in $f$, with $+1$ if $v\in T'_p$ and $-1$ otherwise;
+\item $(\mathbf{z}_v,\mathbf{y}_{p,z}, \pm1)$ if it is not saturated in $f$, with $+1$ if $v\notin T'_p$ and $-1$ otherwise.
+\end{itemize}
+To summarize, arcs are oriented left to right if they correspond to a presence of flow in $f$, and right to left if they correspond to an absence of flow. They are positively weighted if we want them to stay at their current state, and negatively if we want them to switch. Let us compute the weight of such graph.
+
+\begin{multiline*}
+	w(G_f)  = \sum_{e\in E_f} w(e_f)  \\
+	= 
+	(\alpha - \beta -\gamma) N_1 + (\alpha +\beta - \gamma) N_2 + (\alpha+\beta+\gamma) N_3 
+	\\ + 
+	\#V\times N - 4 \sum_p 3-\#(T_p\cap T'_p) \\ 
+	=(\#V-12+\alpha-\beta-\gamma)\times N + 4Q_V + 2\beta N_2 + 2(\beta+\gamma) N_3 \\
+\end{multiline*}
+
+As for the mode 3-strict, one can check that the difference of two such graphs corresponding to the same $(n_v)$ is always eulerian. Hence we can navigate in this class with the same greedy algorithm that discovers positive cycles and flips them.
+
+The function that we optimize is 
+$$
+2Q_V +  \beta N_2 + (\beta+\gamma) N_3.
+$$
+The choice of parameters $\beta$ and $\gamma$ should be lead by the following question: For $\beta$, where to put the tradeoff between zone dispersion and distance to the previous configuration? For $\gamma$, do we prefer to have more partitions spread between 2 zones, or have less between at least 2 zones but more between 3 zones.
+
+The quantity $Q_V$ varies between $0$ and $3N$, it should be of order $N$. The quantity $N_2+N_3$ should also be of order $N$ (it is exactly $N$ in the strict mode). So the two terms of the function are comparable. 
+
+
+\bibliography{optimal_layout} 
+\bibliographystyle{ieeetr}
+
+\end{document}
+
+
+
@@ -0,0 +1,10 @@
+*
+
+!*.txt
+!*.md
+!*.tex
+
+!talk.pdf
+!Makefile
+!.gitignore
+
@@ -0,0 +1,10 @@
+ASSETS=../assets/deuxfleurs.pdf
+
+talk.pdf: talk.tex $(ASSETS)
+	pdflatex talk.tex
+
+assets/%.pdf: assets/%.svg
+	inkscape -D -z --file=$^ --export-pdf=$@
+
+assets/%.pdf_tex: assets/%.svg
+	inkscape -D -z --file=$^ --export-pdf=$@ --export-latex
@@ -0,0 +1,370 @@
+\nonstopmode
+\documentclass[aspectratio=169]{beamer}
+\usepackage[utf8]{inputenc}
+% \usepackage[frenchb]{babel}
+\usepackage{amsmath}
+\usepackage{mathtools}
+\usepackage{breqn}
+\usepackage{multirow}
+\usetheme{boxes}
+\usepackage{graphicx}
+\usepackage{import}
+\usepackage{adjustbox}
+%\useoutertheme[footline=authortitle,subsection=false]{miniframes}
+%\useoutertheme[footline=authorinstitute,subsection=false]{miniframes}
+\useoutertheme{infolines}
+\setbeamertemplate{headline}{}
+
+\beamertemplatenavigationsymbolsempty
+
+\definecolor{TitleOrange}{RGB}{255,137,0}
+\setbeamercolor{title}{fg=TitleOrange}
+\setbeamercolor{frametitle}{fg=TitleOrange}
+
+\definecolor{ListOrange}{RGB}{255,145,5}
+\setbeamertemplate{itemize item}{\color{ListOrange}$\blacktriangleright$}
+
+\definecolor{verygrey}{RGB}{70,70,70}
+\setbeamercolor{normal text}{fg=verygrey}
+
+
+\usepackage{tabu}
+\usepackage{multicol}
+\usepackage{vwcol}
+\usepackage{stmaryrd}
+\usepackage{graphicx}
+
+\usepackage[normalem]{ulem}
+
+\AtBeginSection[]{
+  \begin{frame}
+  \vfill
+  \centering
+  \begin{beamercolorbox}[sep=8pt,center,shadow=true,rounded=true]{title}
+    \usebeamerfont{title}\insertsectionhead\par%
+  \end{beamercolorbox}
+  \vfill
+  \end{frame}
+}
+
+\title{Garage}
+\subtitle{a lightweight and robust geo-distributed data storage system}
+\author{Alex Auvolat, Deuxfleurs}
+\date{SEED webinar, 2024-01-12}
+
+\begin{document}
+
+% \begin{frame}
+% 	\centering
+% 	\includegraphics[width=.3\linewidth]{../../sticker/Garage.png}
+% 	\vspace{1em}
+% 
+% 	{\large\bf Alex Auvolat, Deuxfleurs Association}
+% 	\vspace{1em}
+% 
+% 	\url{https://garagehq.deuxfleurs.fr/}
+% 
+% 	%Matrix channel: \texttt{\#garage:deuxfleurs.fr}
+% \end{frame}
+
+\begin{frame}
+	%\frametitle{Who I am}
+	\begin{columns}[t]
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.4\linewidth, valign=t]{../assets/alex.jpg}
+		\end{column}
+		\begin{column}{.6\textwidth}
+			\textbf{Alex Auvolat}\\
+			Member of Deuxfleurs, lead developer of Garage
+		\end{column}
+		\begin{column}{.2\textwidth}
+			~
+		\end{column}
+	\end{columns}
+	\vspace{.5em}
+
+	\begin{columns}[t]
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.6\linewidth, valign=t]{../../logo/garage-notext.png}
+		\end{column}
+		\begin{column}{.6\textwidth}
+			\\\textbf{Garage}\\
+			A self-hosted alternative to S3 for object storage
+		\end{column}
+		\begin{column}{.2\textwidth}
+			~
+		\end{column}
+	\end{columns}
+	\vspace{2em}
+
+	\begin{columns}[t]
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.5\linewidth, valign=t]{../assets/deuxfleurs.pdf}
+		\end{column}
+		\begin{column}{.6\textwidth}
+			\textbf{Deuxfleurs}\\
+			A non-profit self-hosting collective,\\
+			member of the CHATONS network
+		\end{column}
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.7\linewidth, valign=t]{../assets/logo_chatons.png}
+		\end{column}
+	\end{columns}
+
+\end{frame}
+
+\begin{frame}
+	\frametitle{Stable vs Resilient}
+
+	\hspace{1em}
+	\begin{minipage}{7cm}
+		\textbf{Building a "stable" system:}
+		\vspace{1em}
+
+		Enterprise-grade systems typically employ:
+		\vspace{1em}
+		\begin{itemize}
+			\item RAID
+			\item Redundant power grid + UPS
+			\item Redundant Internet connections
+			\item Low-latency links
+			\item ... 
+		\end{itemize}
+		\vspace{1em}
+		$\to$ costly, only worth at DC scale\\
+		$\to$ still risk of DC-level incident...
+	\end{minipage}
+	\hfill
+	\begin{minipage}{7cm}
+		\textbf{Building a \underline{resilient} system:}
+		\vspace{1em}
+
+		An alternative, cheaper way:
+		\vspace{1em}
+		\begin{itemize}
+			\item Commodity hardware \\(e.g. old desktop PCs)
+				\vspace{.5em}
+			\item Commodity Internet \\(e.g. FTTB, FTTH) and power grid
+				\vspace{.5em}
+			\item \textbf{Geographical redundancy} \\(multi-site replication)
+		\end{itemize}
+		\vspace{1.5em}
+	\end{minipage}
+	\hspace{1em}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Example: our infrastructure at Deuxfleurs}
+	\only<1>{
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{../assets/neptune.jpg}
+		\end{center}
+	}
+	\only<2>{
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{../assets/atuin.jpg}
+		\end{center}
+	}
+	\only<3>{
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{../assets/inframap_jdll2023.pdf}
+		\end{center}
+	}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Object storage: simpler than file systems}
+
+	\begin{minipage}{6cm}
+		Only two operations:
+		\vspace{1em}
+		\begin{itemize}
+			\item Put an object at a key
+				\vspace{1em}
+			\item Retrieve an object from its key
+		\end{itemize}
+		\vspace{1em}
+		{\footnotesize (and a few others)}
+
+		\vspace{1em}
+		Sufficient for many applications!
+	\end{minipage}
+	\hfill
+	\begin{minipage}{8cm}
+		\begin{center}
+			\vspace{2em}
+			\includegraphics[height=6em]{../2020-12-02_wide-team/img/Amazon-S3.jpg}
+			\hspace{2em}
+			\includegraphics[height=5em]{../assets/minio.png}
+
+			\vspace{2em}
+			\includegraphics[height=6em]{../../logo/garage_hires_crop.png}
+		\end{center}
+		\vspace{1em}
+	\end{minipage}
+\end{frame}
+
+
+\begin{frame}
+	\frametitle{The data model of object storage}
+	Object storage is basically a key-value store:
+	\vspace{1em}
+
+		\begin{center}
+		\begin{tabular}{|l|p{8cm}|}
+			\hline
+			\textbf{Key: file path + name} & \textbf{Value: file data + metadata} \\
+			\hline
+			\hline
+			\texttt{index.html} &
+				\texttt{Content-Type: text/html; charset=utf-8} \newline
+				\texttt{Content-Length: 24929} \newline
+				\texttt{<binary blob>} \\ 
+			\hline
+			\texttt{img/logo.svg} &
+				\texttt{Content-Type: text/svg+xml} \newline
+				\texttt{Content-Length: 13429} \newline
+				\texttt{<binary blob>} \\ 
+			\hline
+			\texttt{download/index.html} &
+				\texttt{Content-Type: text/html; charset=utf-8} \newline
+				\texttt{Content-Length: 26563} \newline
+				\texttt{<binary blob>} \\ 
+			\hline
+		\end{tabular}
+		\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Implementation: consensus vs weak consistency}
+
+	\hspace{1em}
+	\begin{minipage}{7cm}
+		\textbf{Consensus-based systems:}
+		\vspace{1em}
+		\begin{itemize}
+			\item \textbf{Leader-based:} a leader is elected to coordinate
+				all reads and writes
+				\vspace{1em}
+			\item Allows for \textbf{sequential reasoning}:
+				program as if running on a single machine
+				\vspace{1em}
+			\item Serializability is one of the \\
+				\textbf{strongest consistency guarantees}
+				\vspace{1em}
+			\item \textbf{Costly}, the leader is a bottleneck;
+				leader elections on failure take time
+		\end{itemize}
+	\end{minipage}
+	\hfill
+	\begin{minipage}{7cm} \visible<2->{
+		\textbf{Weakly consistent systems:}
+		\vspace{1em}
+		\begin{itemize}
+			\item \textbf{Nodes are equivalent}, any node
+				can originate a read or write operation
+				\vspace{1em}
+			\item \textbf{Operations must be independent},
+				conflicts are resolved after the fact
+				\vspace{1em}
+			\item Strongest achievable consistency:\\
+				\textbf{read-after-write consistency}\\(using quorums)
+				\vspace{1em}
+			\item \textbf{Fast}, no single bottleneck;\\
+				works transparently with offline nodes
+		\end{itemize}
+	} \end{minipage}
+	\hspace{1em}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Why avoid consensus?}
+	Consensus can be implemented reasonably well in practice, so why avoid it?
+	\vspace{2em}
+	\begin{itemize}
+		\item \textbf{Software complexity:} RAFT and PAXOS are complex beasts;\\
+			harder to prove, harder to reason about
+			\vspace{1.5em}
+		\item \textbf{Performance issues:}
+			\vspace{1em}
+			\begin{itemize}
+				\item Taking a decision may take an \textbf{arbitrary number of steps} (in adverse scenarios)
+					\vspace{1em}
+				\item The leader is a \textbf{bottleneck} for all requests;\\
+					even in leaderless approaches, \textbf{all nodes must process all operations in order}
+					\vspace{1em}
+				\item Particularly \textbf{sensitive to higher latency} between nodes
+			\end{itemize}
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Objective: the right level of consistency for Garage}
+
+	\underline{Constraints:} slow network (geographical distance), node unavailability/crashes\\
+	\underline{Objective:} maximize availability, maintain an \emph{appropriate level of consistency}\\
+	\vspace{1em}
+	\begin{enumerate}
+		\item<2-> \textbf{Weak consistency for most things}\\
+			\vspace{1em}
+			\underline{Example:} \texttt{PutObject}\\
+			\vspace{.5em}
+			If two clients write the same
+			object at the same time, one of the two is implicitly overwritten.
+			No need to coordinate, use a \emph{last-writer-wins register}.
+			\vspace{1em}
+		\item<3-> \textbf{Stronger consistency only when necessary}\\
+			\vspace{1em}
+			\underline{Example:} \texttt{CreateBucket}\\
+			\vspace{.5em}
+			A bucket is a reserved name in a shared namespace,
+			two clients should be prevented from both creating the same bucket
+			(\emph{mutual exclusion}).
+	\end{enumerate}
+\end{frame}
+
+\begin{frame}
+	\frametitle{The possibility of \emph{leaderless consensus}}
+	Currently, Garage \emph{only has weak consistency}. Is fast, but \texttt{CreateBucket} is broken!
+
+	\visible<2->{
+		\vspace{1em}
+		Leaderless consensus (Antoniadis et al., 2023) alleviates issues with RAFT and PAXOS:
+			\vspace{1em}
+		\begin{itemize}
+			\item \textbf{No leader.} All nodes participate equally at each time step,
+				and different nodes can be unavailable at different times without issues.
+				\\ \vspace{.5em} $\to$ better tolerance to the high latency (remove bottleneck issue)
+				\\ $\to$ tolerates crash transparently
+				\vspace{1em}
+			\item \textbf{Simpler formalization.} The algorithm is very simple to express and to analyze in mathematical terms.
+		\end{itemize}
+	}
+	\visible<3->{
+		\vspace{1em}
+		One of the possible subjects for this PhD:
+		\\$\to$ \emph{integration of leaderless consensus in Garage} + testing + perf eval, etc.
+	}
+\end{frame}
+
+\begin{frame}
+	\begin{center}
+			\includegraphics[width=.25\linewidth]{../../logo/garage_hires.png}\\
+			\vspace{-1em}
+		\url{https://garagehq.deuxfleurs.fr/}\\
+		\url{mailto:garagehq@deuxfleurs.fr}\\
+		\texttt{\#garage:deuxfleurs.fr} on Matrix
+
+		\vspace{1.5em}
+			\includegraphics[width=.06\linewidth]{../assets/rust_logo.png}
+			\includegraphics[width=.13\linewidth]{../assets/AGPLv3_Logo.png}
+	\end{center}
+\end{frame}
+
+\end{document}
+
+%% vim: set ts=4 sw=4 tw=0 noet spelllang=en :
@@ -0,0 +1,17 @@
+*
+
+!*.txt
+!*.md
+
+!assets
+
+!.gitignore
+!*.svg
+!*.png
+!*.jpg
+!*.tex
+!Makefile
+!.gitignore
+!assets/*.drawio.pdf
+
+!talk.pdf
@@ -0,0 +1,19 @@
+ASSETS=../assets/lattice/lattice1.pdf_tex \
+	   ../assets/lattice/lattice2.pdf_tex \
+	   ../assets/lattice/lattice3.pdf_tex \
+	   ../assets/lattice/lattice4.pdf_tex \
+	   ../assets/lattice/lattice5.pdf_tex \
+	   ../assets/lattice/lattice6.pdf_tex \
+	   ../assets/lattice/lattice7.pdf_tex \
+	   ../assets/lattice/lattice8.pdf_tex \
+	   ../assets/logos/deuxfleurs.pdf \
+	   ../assets/timeline-22-24.pdf
+
+talk.pdf: talk.tex $(ASSETS)
+	pdflatex talk.tex
+
+%.pdf: %.svg
+	inkscape -D -z --file=$^ --export-pdf=$@
+
+%.pdf_tex: %.svg
+	inkscape -D -z --file=$^ --export-pdf=$@ --export-latex
@@ -0,0 +1,764 @@
+\nonstopmode
+\documentclass[aspectratio=169,xcolor={svgnames}]{beamer}
+\usepackage[utf8]{inputenc}
+% \usepackage[frenchb]{babel}
+\usepackage{amsmath}
+\usepackage{mathtools}
+\usepackage{breqn}
+\usepackage{multirow}
+\usetheme{boxes}
+\usepackage{graphicx}
+\usepackage{import}
+\usepackage{adjustbox}
+\usepackage[absolute,overlay]{textpos}
+%\useoutertheme[footline=authortitle,subsection=false]{miniframes}
+%\useoutertheme[footline=authorinstitute,subsection=false]{miniframes}
+\useoutertheme{infolines}
+\setbeamertemplate{headline}{}
+
+\beamertemplatenavigationsymbolsempty
+
+\definecolor{TitleOrange}{RGB}{255,137,0}
+\setbeamercolor{title}{fg=TitleOrange}
+\setbeamercolor{frametitle}{fg=TitleOrange}
+
+\definecolor{ListOrange}{RGB}{255,145,5}
+\setbeamertemplate{itemize item}{\color{ListOrange}$\blacktriangleright$}
+
+\definecolor{verygrey}{RGB}{70,70,70}
+\setbeamercolor{normal text}{fg=verygrey}
+
+
+\usepackage{tabu}
+\usepackage{multicol}
+\usepackage{vwcol}
+\usepackage{stmaryrd}
+\usepackage{graphicx}
+
+\usepackage[normalem]{ulem}
+
+\AtBeginSection[]{
+  \begin{frame}
+  \vfill
+  \centering
+  \begin{beamercolorbox}[sep=8pt,center,shadow=true,rounded=true]{title}
+    \usebeamerfont{title}\insertsectionhead\par%
+  \end{beamercolorbox}
+  \vfill
+  \end{frame}
+}
+
+\title{Garage, the low-tech storage platform for geo-distributed clusters}
+\author{Alex Auvolat, Deuxfleurs}
+\date{FOSDEM'24, 2024-02-03}
+
+\begin{document}
+
+\begin{frame}
+	\centering
+	\includegraphics[width=.3\linewidth]{../../sticker/Garage.png}
+	\vspace{1em}
+
+	{\large\bf Alex Auvolat, Deuxfleurs Association}
+	\vspace{1em}
+
+	\url{https://garagehq.deuxfleurs.fr/}
+
+	Matrix channel: \texttt{\#garage:deuxfleurs.fr}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Who I am}
+	\begin{columns}[t]
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.4\linewidth, valign=t]{../assets/alex.jpg}
+		\end{column}
+		\begin{column}{.6\textwidth}
+			\textbf{Alex Auvolat}\\
+			PhD; co-founder of Deuxfleurs
+		\end{column}
+		\begin{column}{.2\textwidth}
+			~
+		\end{column}
+	\end{columns}
+	\vspace{2em}
+
+	\begin{columns}[t]
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.5\linewidth, valign=t]{../assets/logos/deuxfleurs.pdf}
+		\end{column}
+		\begin{column}{.6\textwidth}
+			\textbf{Deuxfleurs}\\
+			A non-profit self-hosting collective,\\
+			member of the CHATONS network
+		\end{column}
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.7\linewidth, valign=t]{../assets/logos/logo_chatons.png}
+		\end{column}
+	\end{columns}
+
+\end{frame}
+
+\begin{frame}
+	\frametitle{Our objective at Deuxfleurs}
+	
+	\begin{center}
+		\textbf{Promote self-hosting and small-scale hosting\\
+			as an alternative to large cloud providers}
+	\end{center}
+	\vspace{2em}
+	\visible<2->{
+		Why is it hard?
+	}
+	\visible<3->{
+		\vspace{2em}
+		\begin{center}
+			\textbf{\underline{Resilience}}\\
+			{\footnotesize we want good uptime/availability with low supervision}
+		\end{center}
+	}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Building a resilient system with cheap stuff}
+
+	\only<1,4-7>{
+		\begin{itemize}
+			\item \textcolor<5->{gray}{Commodity hardware (e.g. old desktop PCs)\\
+				\vspace{.5em}
+				\visible<4->{{\footnotesize (can die at any time)}}}
+				\vspace{1.5em}
+			\item<5-> \textcolor<7->{gray}{Regular Internet (e.g. FTTB, FTTH) and power grid connections\\
+				\vspace{.5em}
+				\visible<6->{{\footnotesize (can be unavailable randomly)}}}
+				\vspace{1.5em}
+			\item<7-> \textbf{Geographical redundancy} (multi-site replication)
+		\end{itemize}
+	}
+	\only<2>{
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{../assets/neptune.jpg}
+		\end{center}
+	}
+	\only<3>{
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{../assets/atuin.jpg}
+		\end{center}
+	}
+	\only<8>{
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{../assets/inframap_jdll2023.pdf}
+		\end{center}
+	}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Object storage: a crucial component}
+		\begin{center}
+			\includegraphics[height=6em]{../assets/logos/Amazon-S3.jpg}
+			\hspace{3em}
+			\visible<2->{\includegraphics[height=5em]{../assets/logos/minio.png}}
+			\hspace{3em}
+			\visible<3>{\includegraphics[height=6em]{../../logo/garage_hires_crop.png}}
+		\end{center}
+		\vspace{1em}
+	S3: a de-facto standard, many compatible applications
+
+	\vspace{1em}
+	\visible<2->{MinIO is self-hostable but not suited for geo-distributed deployments}
+
+	\vspace{1em}
+	\visible<3->{\textbf{Garage is a self-hosted drop-in replacement for the Amazon S3 object store}}
+\end{frame}
+
+\begin{frame}
+	\frametitle{CRDTs / weak consistency instead of consensus}
+
+	\underline{Internally, Garage uses only CRDTs} (conflict-free replicated data types)
+
+	\vspace{2em}
+	Why not Raft, Paxos, ...? Issues of consensus algorithms:
+
+	\vspace{1em}
+	\begin{itemize}
+		\item<2-> \textbf{Software complexity}
+			\vspace{1em}
+		\item<3-> \textbf{Performance issues:}
+			\vspace{.5em}
+			\begin{itemize}
+				\item<4-> The leader is a \textbf{bottleneck} for all requests\\
+					\vspace{.5em}
+				\item<5-> \textbf{Sensitive to higher latency} between nodes
+					\vspace{.5em}
+				\item<6-> \textbf{Takes time to reconverge} when disrupted (e.g. node going down)
+			\end{itemize}
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{The data model of object storage}
+	Object storage is basically a \textbf{key-value store}:
+	\vspace{.5em}
+
+	{\scriptsize
+		\begin{center}
+		\begin{tabular}{|l|p{7cm}|}
+			\hline
+			\textbf{Key: file path + name} & \textbf{Value: file data + metadata} \\
+			\hline
+			\hline
+			\texttt{index.html} &
+				\texttt{Content-Type: text/html; charset=utf-8} \newline
+				\texttt{Content-Length: 24929} \newline
+				\texttt{<binary blob>} \\ 
+			\hline
+			\texttt{img/logo.svg} &
+				\texttt{Content-Type: text/svg+xml} \newline
+				\texttt{Content-Length: 13429} \newline
+				\texttt{<binary blob>} \\ 
+			\hline
+			\texttt{download/index.html} &
+				\texttt{Content-Type: text/html; charset=utf-8} \newline
+				\texttt{Content-Length: 26563} \newline
+				\texttt{<binary blob>} \\ 
+			\hline
+		\end{tabular}
+		\end{center}
+		}
+
+	\vspace{1em}
+	\begin{itemize}
+		\item<2> Maps well to CRDT data types
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Performance gains in practice}
+	\begin{center}
+		\includegraphics[width=.8\linewidth]{../assets/perf/endpoint_latency_0.7_0.8_minio.png}
+	\end{center}
+\end{frame}
+
+
+% ======================================== TIMELINE
+% ======================================== TIMELINE
+% ======================================== TIMELINE
+
+\section{Recent developments}
+
+% ====================== v0.7.0 ===============================
+
+\begin{frame}
+	\begin{center}
+		\includegraphics[width=.8\linewidth]{../assets/timeline-22-24.pdf}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{April 2022 - Garage v0.7.0}
+	Focus on \underline{observability and ecosystem integration}
+	\vspace{2em}
+	\begin{itemize}
+		\item \textbf{Monitoring:} metrics and traces, using OpenTelemetry
+			\vspace{1em}
+		\item Replication modes with 1 or 2 copies / weaker consistency
+			\vspace{1em}
+		\item Kubernetes integration for node discovery
+			\vspace{1em}
+		\item Admin API (v0.7.2)
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Metrics (Prometheus + Grafana)}
+	\begin{center}
+		\includegraphics[width=.9\linewidth]{../assets/screenshots/grafana_dashboard.png}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Traces (Jaeger)}
+	\begin{center}
+		\includegraphics[width=.8\linewidth]{../assets/screenshots/jaeger_listobjects.png}
+	\end{center}
+\end{frame}
+
+% ====================== v0.8.0 ===============================
+
+\begin{frame}
+	\begin{center}
+		\includegraphics[width=.8\linewidth]{../assets/timeline-22-24.pdf}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{November 2022 - Garage v0.8.0}
+	Focus on \underline{performance}
+	\vspace{2em}
+	\begin{itemize}
+		\item \textbf{Alternative metadata DB engines} (LMDB, Sqlite)
+			\vspace{1em}
+		\item \textbf{Performance improvements:} block streaming, various optimizations...
+			\vspace{1em}
+		\item Bucket quotas (max size, max \#objects)
+			\vspace{1em}
+		\item Quality of life improvements, observability, etc.
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{About metadata DB engines}
+	\textbf{Issues with Sled:}
+	\vspace{1em}
+	\begin{itemize}
+		\item Huge files on disk
+			\vspace{.5em}
+		\item Unpredictable performance, especially on HDD
+			\vspace{.5em}
+		\item API limitations
+			\vspace{.5em}
+		\item Not actively maintained
+	\end{itemize}
+
+	\vspace{2em}
+	\textbf{LMDB:} very stable, good performance, file size is reasonable\\
+	\textbf{Sqlite} also available as a second choice
+
+	\vspace{1em}
+	Sled will be removed in Garage v1.0
+\end{frame}
+
+\begin{frame}
+	\frametitle{DB engine performance comparison}
+	\begin{center}
+		\includegraphics[width=.6\linewidth]{../assets/perf/db_engine.png}
+	\end{center}
+	NB: Sqlite was slow due to synchronous mode, now configurable
+\end{frame}
+
+\begin{frame}
+	\frametitle{Block streaming}
+	\begin{center}
+		\only<1>{\includegraphics[width=.8\linewidth]{../assets/schema-streaming-1.png}}
+		\only<2>{\includegraphics[width=.8\linewidth]{../assets/schema-streaming-2.png}}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{TTFB benchmark}
+	\begin{center}
+		\includegraphics[width=.8\linewidth]{../assets/perf/ttfb.png}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Throughput benchmark}
+	\begin{center}
+		\includegraphics[width=.7\linewidth]{../assets/perf/io-0.7-0.8-minio.png}
+	\end{center}
+\end{frame}
+
+% ====================== v0.9.0 ===============================
+
+\begin{frame}
+	\begin{center}
+		\includegraphics[width=.8\linewidth]{../assets/timeline-22-24.pdf}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{October 2023 - Garage v0.9.0}
+	Focus on \underline{streamlining \& usability}
+	\vspace{2em}
+	\begin{itemize}
+		\item Support multiple HDDs per node
+			\vspace{1em}
+		\item S3 compatibility: 
+			\vspace{1em}
+			\begin{itemize}
+				\item support basic lifecycle configurations
+					\vspace{.5em}
+				\item allow for multipart upload part retries
+			\end{itemize}
+			\vspace{1em}
+		\item LMDB by default, deprecation of Sled
+			\vspace{1em}
+		\item New layout computation algorithm
+	\end{itemize}
+\end{frame}
+
+
+\begin{frame}
+	\frametitle{Layout computation}
+	\begin{overprint}
+		\onslide<1>
+		\begin{center}
+			\includegraphics[width=\linewidth, trim=0 0 0 -4cm]{../assets/screenshots/garage_status_0.9_prod_zonehl.png}
+		\end{center}
+		\onslide<2>
+		\begin{center}
+			\includegraphics[width=.7\linewidth]{../assets/map.png}
+		\end{center}
+	\end{overprint}
+	\vspace{1em}
+	Garage stores replicas on different zones when possible
+\end{frame}
+
+\begin{frame}
+	\frametitle{What a "layout" is}
+		\textbf{A layout is a precomputed index table:}
+		\vspace{1em}
+
+		{\footnotesize
+			\begin{center}
+			\begin{tabular}{|l|l|l|l|}
+				\hline
+				\textbf{Partition} & \textbf{Node 1} & \textbf{Node 2} & \textbf{Node 3} \\
+				\hline
+				\hline
+				Partition 0 & df-ymk (bespin) & Abricot (scorpio) & Courgette (neptune) \\
+				\hline
+				Partition 1 & Ananas (scorpio) & Courgette (neptune) & df-ykl (bespin) \\
+				\hline
+				Partition 2 & df-ymf (bespin) & Celeri (neptune) & Abricot (scorpio) \\
+				\hline
+				\hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ \\
+				\hline
+				Partition 255 & Concombre (neptune) & df-ykl (bespin) & Abricot (scorpio) \\
+				\hline
+			\end{tabular}
+			\end{center}
+		}
+
+	\vspace{2em}
+	\visible<2->{
+		The index table is built centrally using an optimal algorithm,\\
+		then propagated to all nodes
+	}
+
+	\vspace{1em}
+	\visible<3->{
+		\footnotesize
+		Oulamara, M., \& Auvolat, A. (2023). \emph{An algorithm for geo-distributed and redundant storage in Garage}.\\ arXiv preprint arXiv:2302.13798.
+	}
+\end{frame}
+
+
+
+% ====================== v0.10.0 ===============================
+
+\begin{frame}
+	\begin{center}
+		\includegraphics[width=.8\linewidth]{../assets/timeline-22-24.pdf}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{October 2023 - Garage v0.10.0 beta}
+	Focus on \underline{consistency}
+	\vspace{2em}
+	\begin{itemize}
+		\item Fix consistency issues when reshuffling data
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Working with weak consistency}
+	Not using consensus limits us to the following:
+	\vspace{2em}
+	\begin{itemize}
+		\item<2-> \textbf{Conflict-free replicated data types} (CRDT)\\
+			\vspace{1em}
+			{\footnotesize Non-transactional key-value stores such as S3 are equivalent to a simple CRDT:\\
+			a map of \textbf{last-writer-wins registers} (each key is its own CRDT)}
+			\vspace{1.5em}
+		\item<3-> \textbf{Read-after-write consistency}\\
+			\vspace{1em}
+			{\footnotesize Can be implemented using quorums on read and write operations}
+	\end{itemize}
+\end{frame}
+
+\begin{frame}[t]
+	\frametitle{CRDT read-after-write consistency using quorums}
+
+	\vspace{1em}
+	{\small
+		\textbf{Property:} If client 1 did an operation $write(x)$ and received an OK response,\\
+			\hspace{2cm} and client 2 starts an operation $read()$ after client 1 received OK,\\
+			\hspace{2cm} then client 2 will read a value $x' \sqsupseteq x$.
+	}
+
+	\vspace{1.5em}
+	\begin{overprint}
+		\onslide<2-9>
+		\begin{figure}
+			\centering
+			\footnotesize
+			\def\svgwidth{.7\textwidth}
+			\only<2>{\import{../assets/lattice/}{lattice1.pdf_tex}}%
+			\only<3>{\import{../assets/lattice/}{lattice2.pdf_tex}}%
+			\only<4>{\import{../assets/lattice/}{lattice3.pdf_tex}}%
+			\only<5>{\import{../assets/lattice/}{lattice4.pdf_tex}}%
+			\only<6>{\import{../assets/lattice/}{lattice5.pdf_tex}}%
+			\only<7>{\import{../assets/lattice/}{lattice6.pdf_tex}}%
+			\only<8>{\import{../assets/lattice/}{lattice7.pdf_tex}}%
+			\only<9>{\import{../assets/lattice/}{lattice8.pdf_tex}}%
+		\end{figure}
+		
+		\onslide<10>
+		\begin{minipage}{.10\textwidth}
+			~
+		\end{minipage}
+		\begin{minipage}{.40\textwidth}
+			\footnotesize
+			\textbf{Algorithm $write(x)$:}
+			\begin{enumerate}
+				\item Broadcast $write(x)$ to all nodes
+				\item Wait for $k > n/2$ nodes to reply OK
+				\item Return OK
+			\end{enumerate}
+		\end{minipage}
+		\begin{minipage}{.40\textwidth}
+			\footnotesize
+			\vspace{1em}
+			\textbf{Algorithm $read()$:}
+			\begin{enumerate}
+				\item Broadcast $read()$ to all nodes
+				\item Wait for $k > n/2$ nodes to reply\\
+					with values $x_1, \dots, x_k$
+				\item Return $x_1 \sqcup \dots \sqcup x_k$
+			\end{enumerate}
+		\end{minipage}
+	\end{overprint}
+\end{frame}
+
+\begin{frame}
+	\frametitle{A hard problem: layout changes}
+	\begin{itemize}
+		\item We rely on quorums $k > n/2$ within each partition:\\
+			$$n=3,~~~~~~~k\ge 2$$
+		\item<2-> When rebalancing, the set of nodes responsible for a partition can change:\\
+
+			\vspace{1em}
+			\begin{minipage}{.04\linewidth}~
+			\end{minipage}
+			\begin{minipage}{.40\linewidth}
+			{\tiny
+				\begin{tabular}{|l|l|l|l|}
+					\hline
+					\textbf{Partition} & \textbf{Node 1} & \textbf{Node 2} & \textbf{Node 3} \\
+					\hline
+					\hline
+					Partition 0 & \textcolor{Crimson}{df-ymk}  & Abricot  & \textcolor{Crimson}{Courgette} \\
+					\hline
+					Partition 1 & Ananas  & \textcolor{Crimson}{Courgette} & \textcolor{Crimson}{df-ykl}  \\
+					\hline
+					Partition 2 & \textcolor{Crimson}{df-ymf}  & \textcolor{Crimson}{Celeri} & Abricot  \\
+					\hline
+					\hspace{1em}$\dots$ & \hspace{1em}$\dots$ & \hspace{1em}$\dots$ & \hspace{1em}$\dots$ \\
+					\hline
+				\end{tabular}
+			}
+			\end{minipage}
+			\begin{minipage}{.04\linewidth}
+				$\to$
+			\end{minipage}
+			\begin{minipage}{.40\linewidth}
+			{\tiny
+				\begin{tabular}{|l|l|l|l|}
+					\hline
+					\textbf{Partition} & \textbf{Node 1} & \textbf{Node 2} & \textbf{Node 3} \\
+					\hline
+					\hline
+					Partition 0 & \textcolor{ForestGreen}{Dahlia}  & Abricot  & \textcolor{ForestGreen}{Eucalyptus} \\
+					\hline
+					Partition 1 & Ananas  & \textcolor{ForestGreen}{Euphorbe}  & \textcolor{ForestGreen}{Doradille} \\
+					\hline
+					Partition 2 & \textcolor{ForestGreen}{Dahlia}  & \textcolor{ForestGreen}{Echinops} & Abricot  \\
+					\hline
+					\hspace{1em}$\dots$ & \hspace{1em}$\dots$ & \hspace{1em}$\dots$ & \hspace{1em}$\dots$ \\
+					\hline
+				\end{tabular}
+			}
+			\end{minipage}
+
+			\vspace{2em}
+		\item<3-> During the rebalancing, new nodes don't yet have the data,\\
+			~~~~~~~~~~~~~~~~~~~and old nodes want to get rid of the data to free up space\\
+			\vspace{1.2em}
+			$\to$ risk of inconsistency, \textbf{how to coordinate?}
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Handling layout changes without losing consistency}
+	\begin{minipage}{.55\textwidth}
+		\begin{itemize}
+			\item \textbf{Solution:}\\
+				\vspace{.5em}
+				\begin{itemize}
+					\item keep track of data transfer to new nodes
+						\vspace{.5em}
+					\item use multiple write quorums\\
+						(new nodes + old nodes\\
+						while data transfer is in progress)
+						\vspace{.5em}
+					\item switching reads to new nodes\\
+						only once copy is finished
+				\end{itemize}
+				\vspace{1em}
+			\item \textbf{Implemented} in v0.10
+				\vspace{1em}
+			\item \textbf{Validated} with Jepsen testing
+		\end{itemize}
+	\end{minipage}
+	\begin{minipage}{.23\textwidth}
+		\includegraphics[width=3cm]{../assets/jepsen-0.9.png}\\
+		{\footnotesize Garage v0.9.0}
+	\end{minipage}
+	\begin{minipage}{.2\textwidth}
+		\includegraphics[width=3cm]{../assets/jepsen-0.10.png}\\
+		{\footnotesize Garage v0.10 beta}
+	\end{minipage}
+\end{frame}
+
+% ====================== v0.10.0 ===============================
+
+\begin{frame}
+	\begin{center}
+		\includegraphics[width=.8\linewidth]{../assets/timeline-22-24.pdf}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Towards v1.0...}
+	Focus on \underline{security \& stability}
+	\vspace{2em}
+	\begin{itemize}
+		\item \textbf{Security audit} in progress by Radically Open Security
+			\vspace{1em}
+		\item Misc. S3 features (SSE-C, ...) and compatibility fixes
+			\vspace{1em}
+		\item Improve UX
+			\vspace{1em}
+		\item Fix bugs
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{...and beyond!}
+	\begin{center}
+		\includegraphics[width=.6\linewidth]{../assets/survey_requested_features.png}
+	\end{center}
+\end{frame}
+
+% ======================================== OPERATING
+% ======================================== OPERATING
+% ======================================== OPERATING
+
+
+\section{Operating big Garage clusters}
+
+\begin{frame}
+	\frametitle{Operating Garage}
+	\begin{center}
+		\only<1-2>{
+			\includegraphics[width=.9\linewidth]{../assets/screenshots/garage_status_0.10.png}
+			\\\vspace{1em}
+			\visible<2>{\includegraphics[width=.9\linewidth]{../assets/screenshots/garage_status_unhealthy_0.10.png}}
+		}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Garage's architecture}
+	\begin{center}
+		\only<1>{\includegraphics[width=.45\linewidth]{../assets/garage.drawio.pdf}}%
+		\only<2>{\includegraphics[width=.6\linewidth]{../assets/garage_sync.drawio.pdf}}%
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Digging deeper}
+	\begin{center}
+		\only<1>{\includegraphics[width=.9\linewidth]{../assets/screenshots/garage_stats_0.10.png}}
+		\only<2>{\includegraphics[width=.5\linewidth]{../assets/screenshots/garage_worker_list_0.10.png}}
+		\only<3>{\includegraphics[width=.6\linewidth]{../assets/screenshots/garage_worker_param_0.10.png}}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Potential limitations and bottlenecks}
+	\begin{itemize}
+		\item Global:
+			\begin{itemize}
+				\item Max. $\sim$100 nodes per cluster (excluding gateways)
+			\end{itemize}
+			\vspace{1em}
+		\item Metadata:
+			\begin{itemize}
+				\item One big bucket = bottleneck, object list on 3 nodes only
+			\end{itemize}
+			\vspace{1em}
+		\item Block manager:
+			\begin{itemize}
+				\item Lots of small files on disk
+				\item Processing the resync queue can be slow
+			\end{itemize}
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Deployment advice for very large clusters}
+	\begin{itemize}
+		\item Metadata storage:
+			\begin{itemize}
+				\item ZFS mirror (x2) on fast NVMe
+				\item Use LMDB storage engine
+			\end{itemize}
+			\vspace{.5em}
+		\item Data block storage:
+			\begin{itemize}
+				\item Use Garage's native multi-HDD support
+				\item XFS on individual drives
+				\item Increase block size (1MB $\to$ 10MB, requires more RAM and good networking)
+				\item Tune \texttt{resync-tranquility} and \texttt{resync-worker-count} dynamically
+			\end{itemize}
+			\vspace{.5em}
+		\item Other :
+			\begin{itemize}
+				\item Split data over several buckets
+				\item Use less than 100 storage nodes
+				\item Use gateway nodes
+			\end{itemize}
+			\vspace{.5em}
+	\end{itemize}
+	Our deployments: $< 10$ TB. Some people have done more!
+\end{frame}
+
+
+% ======================================== END
+% ======================================== END
+% ======================================== END
+
+\begin{frame}
+	\frametitle{Where to find us}
+	\begin{center}
+			\includegraphics[width=.25\linewidth]{../../logo/garage_hires.png}\\
+			\vspace{-1em}
+		\url{https://garagehq.deuxfleurs.fr/}\\
+		\url{mailto:garagehq@deuxfleurs.fr}\\
+		\texttt{\#garage:deuxfleurs.fr} on Matrix
+
+		\vspace{1.5em}
+			\includegraphics[width=.06\linewidth]{../assets/logos/rust_logo.png}
+			\includegraphics[width=.13\linewidth]{../assets/logos/AGPLv3_Logo.png}
+	\end{center}
+\end{frame}
+
+\end{document}
+
+%% vim: set ts=4 sw=4 tw=0 noet spelllang=en :
@@ -0,0 +1,17 @@
+*
+
+!*.txt
+!*.md
+
+!assets
+
+!.gitignore
+!*.svg
+!*.png
+!*.jpg
+!*.tex
+!Makefile
+!.gitignore
+!assets/*.drawio.pdf
+
+!talk.pdf
@@ -0,0 +1,10 @@
+ASSETS=../assets/logos/deuxfleurs.pdf
+
+talk.pdf: talk.tex $(ASSETS)
+	pdflatex talk.tex
+
+%.pdf: %.svg
+	inkscape -D -z --file=$^ --export-pdf=$@
+
+%.pdf_tex: %.svg
+	inkscape -D -z --file=$^ --export-pdf=$@ --export-latex
@@ -0,0 +1,543 @@
+\nonstopmode
+\documentclass[aspectratio=169,xcolor={svgnames}]{beamer}
+\usepackage[utf8]{inputenc}
+% \usepackage[frenchb]{babel}
+\usepackage{amsmath}
+\usepackage{mathtools}
+\usepackage{breqn}
+\usepackage{multirow}
+\usetheme{boxes}
+\usepackage{graphicx}
+\usepackage{import}
+\usepackage{adjustbox}
+\usepackage[absolute,overlay]{textpos}
+%\useoutertheme[footline=authortitle,subsection=false]{miniframes}
+%\useoutertheme[footline=authorinstitute,subsection=false]{miniframes}
+\useoutertheme{infolines}
+\setbeamertemplate{headline}{}
+
+\beamertemplatenavigationsymbolsempty
+
+\definecolor{TitleOrange}{RGB}{255,137,0}
+\setbeamercolor{title}{fg=TitleOrange}
+\setbeamercolor{frametitle}{fg=TitleOrange}
+
+\definecolor{ListOrange}{RGB}{255,145,5}
+\setbeamertemplate{itemize item}{\color{ListOrange}$\blacktriangleright$}
+
+\definecolor{verygrey}{RGB}{70,70,70}
+\setbeamercolor{normal text}{fg=verygrey}
+
+
+\usepackage{tabu}
+\usepackage{multicol}
+\usepackage{vwcol}
+\usepackage{stmaryrd}
+\usepackage{graphicx}
+
+\usepackage[normalem]{ulem}
+
+\AtBeginSection[]{
+  \begin{frame}
+  \vfill
+  \centering
+  \begin{beamercolorbox}[sep=8pt,center,shadow=true,rounded=true]{title}
+    \usebeamerfont{title}\insertsectionhead\par%
+  \end{beamercolorbox}
+  \vfill
+  \end{frame}
+}
+
+\title{Garage}
+\author{Alex Auvolat, Deuxfleurs}
+\date{Capitoul, 2024-02-29}
+
+\begin{document}
+
+\begin{frame}
+	\centering
+	\includegraphics[width=.3\linewidth]{../../sticker/Garage.png}
+	\vspace{1em}
+
+	{\large\bf Alex Auvolat, Deuxfleurs Association}
+	\vspace{1em}
+
+	\url{https://garagehq.deuxfleurs.fr/}
+
+	Matrix channel: \texttt{\#garage:deuxfleurs.fr}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Who I am}
+	\begin{columns}[t]
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.4\linewidth, valign=t]{../assets/alex.jpg}
+		\end{column}
+		\begin{column}{.6\textwidth}
+			\textbf{Alex Auvolat}\\
+			PhD; co-founder of Deuxfleurs
+		\end{column}
+		\begin{column}{.2\textwidth}
+			~
+		\end{column}
+	\end{columns}
+	\vspace{2em}
+
+	\begin{columns}[t]
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.5\linewidth, valign=t]{../assets/logos/deuxfleurs.pdf}
+		\end{column}
+		\begin{column}{.6\textwidth}
+			\textbf{Deuxfleurs}\\
+			A non-profit self-hosting collective,\\
+			member of the CHATONS network
+		\end{column}
+		\begin{column}{.2\textwidth}
+			\centering
+			\adjincludegraphics[width=.7\linewidth, valign=t]{../assets/logos/logo_chatons.png}
+		\end{column}
+	\end{columns}
+
+\end{frame}
+
+\begin{frame}
+	\frametitle{Our objective at Deuxfleurs}
+	
+	\begin{center}
+		\textbf{Promote self-hosting and small-scale hosting\\
+			as an alternative to large cloud providers}
+	\end{center}
+	\vspace{2em}
+	\visible<2->{
+		Why is it hard?
+		\vspace{2em}
+		\begin{center}
+			\textbf{\underline{Resilience}}\\
+			{\footnotesize we want good uptime/availability with low supervision}
+		\end{center}
+	}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Our very low-tech infrastructure}
+
+	\only<1,3-6>{
+		\begin{itemize}
+			\item \textcolor<4->{gray}{Commodity hardware (e.g. old desktop PCs)\\
+				\vspace{.5em}
+				\visible<3->{{\footnotesize (can die at any time)}}}
+				\vspace{1.5em}
+			\item<4-> \textcolor<6->{gray}{Regular Internet (e.g. FTTB, FTTH) and power grid connections\\
+				\vspace{.5em}
+				\visible<5->{{\footnotesize (can be unavailable randomly)}}}
+				\vspace{1.5em}
+			\item<6-> \textbf{Geographical redundancy} (multi-site replication)
+		\end{itemize}
+	}
+	\only<2>{
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{../assets/neptune.jpg}
+		\end{center}
+	}
+	\only<7>{
+		\begin{center}
+			\includegraphics[width=.8\linewidth]{../assets/inframap_jdll2023.pdf}
+		\end{center}
+	}
+\end{frame}
+
+\begin{frame}
+	\frametitle{How to make this happen}
+	\begin{center}
+		\only<1>{\includegraphics[width=.8\linewidth]{../assets/intro/slide1.png}}%
+		\only<2>{\includegraphics[width=.8\linewidth]{../assets/intro/slide2.png}}%
+		\only<3>{\includegraphics[width=.8\linewidth]{../assets/intro/slide3.png}}%
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Distributed file systems are slow}
+	File systems are complex, for example:
+	\vspace{1em}
+	\begin{itemize}
+		\item Concurrent modification by several processes
+			\vspace{1em}
+		\item Folder hierarchies
+			\vspace{1em}
+		\item Other requirements of the POSIX spec (e.g.~locks)
+	\end{itemize}
+	\vspace{1em}
+	Coordination in a distributed system is costly
+
+	\vspace{1em}
+	Costs explode with commodity hardware / Internet connections\\
+	{\small (we experienced this!)}
+\end{frame}
+
+\begin{frame}
+	\frametitle{A simpler solution: object storage}
+	Only two operations:
+	\vspace{1em}
+	\begin{itemize}
+		\item Put an object at a key
+			\vspace{1em}
+		\item Retrieve an object from its key
+	\end{itemize}
+	\vspace{1em}
+	{\footnotesize (and a few others)}
+
+	\vspace{1em}
+	Sufficient for many applications!
+\end{frame}
+
+\begin{frame}
+	\frametitle{A simpler solution: object storage}
+		\begin{center}
+			\includegraphics[height=6em]{../assets/logos/Amazon-S3.jpg}
+			\hspace{3em}
+			\visible<2->{\includegraphics[height=5em]{../assets/logos/minio.png}}
+			\hspace{3em}
+			\visible<3>{\includegraphics[height=6em]{../../logo/garage_hires_crop.png}}
+		\end{center}
+		\vspace{1em}
+	S3: a de-facto standard, many compatible applications
+
+	\vspace{1em}
+	\visible<2->{MinIO is self-hostable but not suited for geo-distributed deployments}
+
+	\vspace{1em}
+	\visible<3->{\textbf{Garage is a self-hosted drop-in replacement for the Amazon S3 object store}}
+\end{frame}
+
+% --------- BASED ON CRDTS ----------
+
+\section{Principle 1: based on CRDTs}
+
+\begin{frame}
+	\frametitle{CRDTs / weak consistency instead of consensus}
+
+	\underline{Internally, Garage uses only CRDTs} (conflict-free replicated data types)
+
+	\vspace{2em}
+	Why not Raft, Paxos, ...? Issues of consensus algorithms:
+
+	\vspace{1em}
+	\begin{itemize}
+		\item<2-> \textbf{Software complexity}
+			\vspace{1em}
+		\item<3-> \textbf{Performance issues:}
+			\vspace{.5em}
+			\begin{itemize}
+				\item<4-> The leader is a \textbf{bottleneck} for all requests\\
+					\vspace{.5em}
+				\item<5-> \textbf{Sensitive to higher latency} between nodes
+					\vspace{.5em}
+				\item<6-> \textbf{Takes time to reconverge} when disrupted (e.g. node going down)
+			\end{itemize}
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{The data model of object storage}
+	Object storage is basically a \textbf{key-value store}:
+	\vspace{.5em}
+
+	{\scriptsize
+		\begin{center}
+		\begin{tabular}{|l|p{7cm}|}
+			\hline
+			\textbf{Key: file path + name} & \textbf{Value: file data + metadata} \\
+			\hline
+			\hline
+			\texttt{index.html} &
+				\texttt{Content-Type: text/html; charset=utf-8} \newline
+				\texttt{Content-Length: 24929} \newline
+				\texttt{<binary blob>} \\ 
+			\hline
+			\texttt{img/logo.svg} &
+				\texttt{Content-Type: text/svg+xml} \newline
+				\texttt{Content-Length: 13429} \newline
+				\texttt{<binary blob>} \\ 
+			\hline
+			\texttt{download/index.html} &
+				\texttt{Content-Type: text/html; charset=utf-8} \newline
+				\texttt{Content-Length: 26563} \newline
+				\texttt{<binary blob>} \\ 
+			\hline
+		\end{tabular}
+		\end{center}
+		}
+
+	\vspace{.5em}
+	\begin{itemize}
+		\item<2-> Maps well to CRDT data types
+		\item<3> Read-after-write consistency with quorums
+	\end{itemize}
+\end{frame}
+
+
+\begin{frame}
+	\frametitle{Performance gains in practice}
+	\begin{center}
+		\includegraphics[width=.8\linewidth]{../assets/perf/endpoint_latency_0.7_0.8_minio.png}
+	\end{center}
+\end{frame}
+
+% --------- GEO-DISTRIBUTED MODEL ----------
+
+\section{Principle 2: geo-distributed data model}
+
+\begin{frame}
+	\frametitle{Key-value stores, upgraded: the Dynamo model}
+	\textbf{Two keys:}
+	\begin{itemize}
+		\item Partition key: used to divide data into partitions {\small (a.k.a.~shards)}
+		\item Sort key: used to identify items inside a partition
+	\end{itemize}
+
+	\vspace{1em}
+
+	\begin{center}
+	\begin{tabular}{|l|l|p{3cm}|}
+		\hline
+		\textbf{Partition key: bucket} & \textbf{Sort key: filename} & \textbf{Value} \\
+		\hline
+		\hline
+		\texttt{website} & \texttt{index.html} & (file data) \\
+		\hline
+		\texttt{website} & \texttt{img/logo.svg} & (file data) \\
+		\hline
+		\texttt{website} & \texttt{download/index.html} & (file data) \\
+		\hline
+		\hline
+		\texttt{backup} & \texttt{borg/index.2822} & (file data) \\
+		\hline
+		\texttt{backup} & \texttt{borg/data/2/2329} & (file data) \\
+		\hline
+		\texttt{backup} & \texttt{borg/data/2/2680} & (file data) \\
+		\hline
+		\hline
+		\texttt{private} & \texttt{qq3a2nbe1qjq0ebbvo6ocsp6co} & (file data) \\
+		\hline
+	\end{tabular}
+	\end{center}
+\end{frame}
+
+
+\begin{frame}
+	\frametitle{Layout computation}
+	\begin{overprint}
+		\onslide<1>
+		\begin{center}
+			\includegraphics[width=\linewidth, trim=0 0 0 -4cm]{../assets/screenshots/garage_status_0.9_prod_zonehl.png}
+		\end{center}
+		\onslide<2>
+		\begin{center}
+			\includegraphics[width=.7\linewidth]{../assets/map.png}
+		\end{center}
+	\end{overprint}
+	\vspace{1em}
+	Garage stores replicas on different zones when possible
+\end{frame}
+
+\begin{frame}
+	\frametitle{What a "layout" is}
+		\textbf{A layout is a precomputed index table:}
+		\vspace{1em}
+
+		{\footnotesize
+			\begin{center}
+			\begin{tabular}{|l|l|l|l|}
+				\hline
+				\textbf{Partition} & \textbf{Node 1} & \textbf{Node 2} & \textbf{Node 3} \\
+				\hline
+				\hline
+				Partition 0 & df-ymk (bespin) & Abricot (scorpio) & Courgette (neptune) \\
+				\hline
+				Partition 1 & Ananas (scorpio) & Courgette (neptune) & df-ykl (bespin) \\
+				\hline
+				Partition 2 & df-ymf (bespin) & Celeri (neptune) & Abricot (scorpio) \\
+				\hline
+				\hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ & \hspace{1em}$\vdots$ \\
+				\hline
+				Partition 255 & Concombre (neptune) & df-ykl (bespin) & Abricot (scorpio) \\
+				\hline
+			\end{tabular}
+			\end{center}
+		}
+
+	\vspace{2em}
+	\visible<2->{
+		The index table is built centrally using an optimal algorithm,\\
+		then propagated to all nodes
+	}
+
+	\vspace{1em}
+	\visible<3->{
+		\footnotesize
+		Oulamara, M., \& Auvolat, A. (2023). \emph{An algorithm for geo-distributed and redundant storage in Garage}.\\ arXiv preprint arXiv:2302.13798.
+	}
+\end{frame}
+
+
+\begin{frame}
+	\frametitle{The relationship between \emph{partition} and \emph{partition key}}
+	\begin{center}
+	\begin{tabular}{|l|l|l|l|}
+		\hline
+		\textbf{Partition key} & \textbf{Partition} & \textbf{Sort key} & \textbf{Value} \\
+		\hline
+		\hline
+		\texttt{website} & Partition 12 & \texttt{index.html} & (file data) \\
+		\hline
+		\texttt{website} & Partition 12 & \texttt{img/logo.svg} & (file data) \\
+		\hline
+		\texttt{website} & Partition 12 &\texttt{download/index.html} & (file data) \\
+		\hline
+		\hline
+		\texttt{backup} & Partition 42 & \texttt{borg/index.2822} & (file data) \\
+		\hline
+		\texttt{backup} & Partition 42 & \texttt{borg/data/2/2329} & (file data) \\
+		\hline
+		\texttt{backup} & Partition 42 & \texttt{borg/data/2/2680} & (file data) \\
+		\hline
+		\hline
+		\texttt{private} & Partition 42 & \texttt{qq3a2nbe1qjq0ebbvo6ocsp6co} & (file data) \\
+		\hline
+	\end{tabular}
+	\end{center}
+	\vspace{1em}
+	\textbf{To read or write an item:} hash partition key
+	\\ \hspace{5cm} $\to$ determine partition number (first 8 bits)
+	\\ \hspace{5cm} $\to$ find associated nodes
+\end{frame}
+
+\begin{frame}
+	\frametitle{Garage's internal data structures}
+	\centering
+	\includegraphics[width=.75\columnwidth]{../assets/garage_tables.pdf}
+\end{frame}
+
+% ---------- OPERATING GARAGE ---------
+
+\section{Operating Garage clusters}
+
+\begin{frame}
+	\frametitle{Operating Garage}
+	\begin{center}
+		\only<1-2>{
+			\includegraphics[width=.9\linewidth]{../assets/screenshots/garage_status_0.10.png}
+			\\\vspace{1em}
+			\visible<2>{\includegraphics[width=.9\linewidth]{../assets/screenshots/garage_status_unhealthy_0.10.png}}
+		}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Background synchronization}
+	\begin{center}
+		\includegraphics[width=.6\linewidth]{../assets/garage_sync.drawio.pdf}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Digging deeper}
+	\begin{center}
+		\only<1>{\includegraphics[width=.9\linewidth]{../assets/screenshots/garage_stats_0.10.png}}
+		\only<2>{\includegraphics[width=.5\linewidth]{../assets/screenshots/garage_worker_list_0.10.png}}
+		\only<3>{\includegraphics[width=.6\linewidth]{../assets/screenshots/garage_worker_param_0.10.png}}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Monitoring with Prometheus + Grafana}
+	\begin{center}
+		\includegraphics[width=.9\linewidth]{../assets/screenshots/grafana_dashboard.png}
+	\end{center}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Debugging with traces}
+	\begin{center}
+		\includegraphics[width=.8\linewidth]{../assets/screenshots/jaeger_listobjects.png}
+	\end{center}
+\end{frame}
+
+% ---------- SCALING GARAGE ---------
+
+\section{Scaling Garage clusters}
+
+\begin{frame}
+	\frametitle{Potential limitations and bottlenecks}
+	\begin{itemize}
+		\item Global:
+			\begin{itemize}
+				\item Max. $\sim$100 nodes per cluster (excluding gateways)
+			\end{itemize}
+			\vspace{1em}
+		\item Metadata:
+			\begin{itemize}
+				\item One big bucket = bottleneck, object list on 3 nodes only
+			\end{itemize}
+			\vspace{1em}
+		\item Block manager:
+			\begin{itemize}
+				\item Lots of small files on disk
+				\item Processing the resync queue can be slow
+			\end{itemize}
+	\end{itemize}
+\end{frame}
+
+\begin{frame}
+	\frametitle{Deployment advice for very large clusters}
+	\begin{itemize}
+		\item Metadata storage:
+			\begin{itemize}
+				\item ZFS mirror (x2) on fast NVMe
+				\item Use LMDB storage engine
+			\end{itemize}
+			\vspace{.5em}
+		\item Data block storage:
+			\begin{itemize}
+				\item Use Garage's native multi-HDD support
+				\item XFS on individual drives
+				\item Increase block size (1MB $\to$ 10MB, requires more RAM and good networking)
+				\item Tune \texttt{resync-tranquility} and \texttt{resync-worker-count} dynamically
+			\end{itemize}
+			\vspace{.5em}
+		\item Other :
+			\begin{itemize}
+				\item Split data over several buckets
+				\item Use less than 100 storage nodes
+				\item Use gateway nodes
+			\end{itemize}
+			\vspace{.5em}
+	\end{itemize}
+	Our deployments: $< 10$ TB. Some people have done more!
+\end{frame}
+
+
+% ======================================== END
+% ======================================== END
+% ======================================== END
+
+\begin{frame}
+	\frametitle{Where to find us}
+	\begin{center}
+			\includegraphics[width=.25\linewidth]{../../logo/garage_hires.png}\\
+			\vspace{-1em}
+		\url{https://garagehq.deuxfleurs.fr/}\\
+		\url{mailto:garagehq@deuxfleurs.fr}\\
+		\texttt{\#garage:deuxfleurs.fr} on Matrix
+
+		\vspace{1.5em}
+			\includegraphics[width=.06\linewidth]{../assets/logos/rust_logo.png}
+			\includegraphics[width=.13\linewidth]{../assets/logos/AGPLv3_Logo.png}
+	\end{center}
+\end{frame}
+
+\end{document}
+
+%% vim: set ts=4 sw=4 tw=0 noet spelllang=en :
@@ -0,0 +1,17 @@
+*
+
+!*.txt
+!*.md
+
+!assets
+
+!.gitignore
+!*.svg
+!*.png
+!*.jpg
+!*.tex
+!Makefile
+!.gitignore
+!assets/*.drawio.pdf
+
+!talk.pdf
--- a/Show More
+++ b/Show More