🧑‍💻(devx) prototype OrbStack as local Kubernetes provider on macOS

Add a prototype setup that uses OrbStack as the local Kubernetes
provider on macOS, aiming to save a few GB of RAM compared to the
current stack.

The script has been tested locally but was mostly built through vibe
coding, so it has not been thoroughly reviewed yet.

Follow-ups to look into:

* Simplify the ingress and CoreDNS setup if possible.
* Wire up a way to run tests and lint against the Tilt stack.

Makefile

@@ -389,6 +389,12 @@ build-k8s-cluster: \
     ./bin/start-kind.sh
 .PHONY: build-k8s-cluster
 
+build-k8s-cluster-orbstack: ## setup the kubernetes environment on OrbStack's built-in cluster (macOS)
+build-k8s-cluster-orbstack: \
+    env.d/development/kube-secret
+	./bin/start-orbstack.sh
+.PHONY: build-k8s-cluster-orbstack
+
 start-tilt-keycloak: ## start the kubernetes cluster using kind, without Pro Connect for authentication, use keycloak
 	DEV_ENV=dev-keycloak tilt up --namespace=meet -f ./bin/Tiltfile
 .PHONY: build-k8s-cluster

bin/Tiltfile

@@ -1,5 +1,11 @@
 load('ext://uibutton', 'cmd_button', 'bool_input', 'location')
 load('ext://namespace', 'namespace_create', 'namespace_inject')
+
+# OrbStack's built-in cluster (macOS) is a supported alternative to kind.
+# Recent Tilt versions (>= 0.33) detect it as a local dev cluster; this is
+# a no-op for kind and a safety net for older Tilt versions.
+allow_k8s_contexts('orbstack')
+
 namespace_create('meet')
 
 DEV_ENV = os.getenv('DEV_ENV', 'dev-keycloak')

bin/start-orbstack.sh

@@ -0,0 +1,182 @@
+#!/usr/bin/env bash
+#
+# Bootstrap the local dev environment on OrbStack's built-in Kubernetes
+# cluster (macOS) instead of kind.
+#
+# This replicates what bin/start-kind.sh (numerique-gouv/tools
+# kind/create_cluster.sh) provides, minus what OrbStack makes unnecessary:
+#   - no kind cluster: OrbStack ships a lightweight single-node cluster
+#   - no local registry (kind-registry): OrbStack's cluster shares the
+#     Docker image store, so images built by Tilt are directly visible
+#     to pods. Tilt detects the "orbstack" context as a local cluster
+#     and skips pushing images entirely.
+#
+# Requirements: OrbStack (with Kubernetes enabled), kubectl, mkcert, curl.
+set -o errexit
+
+APPLICATION=${1:-meet}
+CONTEXT="orbstack"
+
+echo "0. Check OrbStack Kubernetes is available"
+if ! command -v mkcert >/dev/null 2>&1; then
+  echo "❌ mkcert is not installed. Install it first: brew install mkcert"
+  exit 1
+fi
+if ! kubectl config get-contexts -o name | grep -qx "${CONTEXT}"; then
+  echo "Context '${CONTEXT}' not found. Trying to start OrbStack Kubernetes..."
+  if command -v orb >/dev/null 2>&1; then
+    orb start k8s
+  else
+    echo "❌ Enable Kubernetes in OrbStack (Settings > Kubernetes) and retry."
+    exit 1
+  fi
+fi
+kubectl config use-context "${CONTEXT}"
+
+echo "0b. Check ports 80/443 are free on localhost"
+# OrbStack forwards LoadBalancer service ports to 127.0.0.1. If the kind
+# cluster is still running, its docker proxy already holds 80/443.
+# Skip the check if ingress-nginx is already installed here: in that case
+# the listener on 80/443 is our own LoadBalancer.
+if ! kubectl -n ingress-nginx get deployment ingress-nginx-controller >/dev/null 2>&1; then
+  for port in 80 443; do
+    if lsof -nP -iTCP:"${port}" -sTCP:LISTEN >/dev/null 2>&1; then
+      echo "❌ Port ${port} is already in use on the host."
+      echo "   If the kind cluster is running, delete it first:"
+      echo "     kind delete cluster --name suite"
+      exit 1
+    fi
+  done
+fi
+
+echo "1. Create ca"
+CURRENT_DIR=$(pwd)
+mkcert -install
+cd /tmp
+mkcert "127.0.0.1.nip.io" "*.127.0.0.1.nip.io"
+cd "${CURRENT_DIR}"
+
+echo "2. Install ingress-nginx (cloud provider: LoadBalancer service)"
+# OrbStack exposes LoadBalancer services on 127.0.0.1, so the cloud
+# manifest replaces kind's hostPort-based deploy. Every sub-step below is
+# guarded individually so the script is safe to re-run after a partial
+# failure (unlike the upstream kind script, which guards the whole block
+# on namespace existence).
+
+# Make sure no stale registry configmap tells Tilt to push to localhost:5001
+# (there is no registry on OrbStack).
+kubectl -n kube-public delete configmap local-registry-hosting --ignore-not-found
+
+if ! kubectl -n ingress-nginx get deployment ingress-nginx-controller >/dev/null 2>&1; then
+  kubectl apply -f https://raw.githubusercontent.com/kubernetes/ingress-nginx/main/deploy/static/provider/cloud/deploy.yaml
+fi
+if ! kubectl -n ingress-nginx get deployment nginx-errors >/dev/null 2>&1; then
+  kubectl apply -n ingress-nginx -f https://raw.githubusercontent.com/kubernetes/ingress-nginx/refs/heads/main/docs/examples/customization/custom-errors/custom-default-backend.yaml
+fi
+kubectl -n ingress-nginx create secret tls mkcert --key /tmp/127.0.0.1.nip.io+1-key.pem --cert /tmp/127.0.0.1.nip.io+1.pem || echo ok
+
+# The meet charts render Ingresses without ingressClassName. The kind
+# provider manifest handles this via --watch-ingress-without-class=true;
+# the cloud manifest does not, so add it here (otherwise: 404 everywhere).
+if ! kubectl -n ingress-nginx get deployment ingress-nginx-controller -o jsonpath='{.spec.template.spec.containers[0].args}' | grep -q 'watch-ingress-without-class'; then
+  kubectl -n ingress-nginx patch deployments.apps ingress-nginx-controller --type 'json' -p '[{"op": "add", "path": "/spec/template/spec/containers/0/args/-", "value":"--watch-ingress-without-class=true"},{"op": "add", "path": "/spec/template/spec/containers/0/args/-", "value":"--default-ssl-certificate=ingress-nginx/mkcert"},{"op": "add", "path": "/spec/template/spec/containers/0/args/-", "value":"--default-backend-service=ingress-nginx/nginx-errors"}
+]'
+fi
+if ! kubectl -n ingress-nginx get deployment nginx-errors -o jsonpath='{.spec.template.spec.containers[0].image}' | grep -q 'error-pages'; then
+  kubectl -n ingress-nginx patch deployment nginx-errors --type=json -p='[
+  {"op": "replace", "path": "/spec/template/spec/containers/0/image", "value": "ghcr.io/tarampampam/error-pages:3.3.0"},
+  {"op": "add", "path": "/spec/template/spec/containers/0/env", "value": [{"name": "TEMPLATE_NAME", "value": "ghost"}, {"name": "SHOW_DETAILS", "value": "false"}, {"name": "SEND_SAME_HTTP_CODE", "value": "true"}]}
+]'
+fi
+cat <<EOF | kubectl apply -n ingress-nginx -f -
+  apiVersion: v1
+  data:
+    allow-snippet-annotations: "true"
+    annotations-risk-level: Critical
+    custom-http-errors: 500,501,502,503,504
+  kind: ConfigMap
+  metadata:
+    name: ingress-nginx-controller
+    namespace: ingress-nginx
+EOF
+
+echo "2b. Wait for the ingress controller to be ready"
+kubectl -n ingress-nginx rollout status deployment/ingress-nginx-controller --timeout=180s
+
+echo "3. Patch CoreDNS so in-cluster pods resolve *.127.0.0.1.nip.io to the ingress"
+# nip.io resolves to 127.0.0.1, which inside a pod is the pod itself.
+# Rewrite these names to the ingress-nginx service, like the kind setup does.
+# Unlike kind, we amend OrbStack's existing Corefile instead of replacing it.
+if ! kubectl -n kube-system get configmap coredns -o jsonpath='{.data.Corefile}' | grep -q '127\.0\.0\.1\.nip\.io'; then
+  kubectl -n kube-system get configmap coredns -o jsonpath='{.data.Corefile}' \
+    | awk '/forward \./ && !done { print "    rewrite stop {"; print "      name regex (.*).127.0.0.1.nip.io ingress-nginx-controller.ingress-nginx.svc.cluster.local answer auto"; print "    }"; done=1 } { print }' \
+    >/tmp/Corefile.orbstack
+  kubectl -n kube-system create configmap coredns --from-file=Corefile=/tmp/Corefile.orbstack --dry-run=client -o yaml | kubectl apply -f -
+  kubectl -n kube-system rollout restart deployments/coredns
+fi
+
+if ! kubectl get ns "${APPLICATION}" >/dev/null 2>&1; then
+  echo "4. Setup namespace"
+  kubectl create ns "${APPLICATION}"
+fi
+kubectl config set-context --current --namespace="${APPLICATION}"
+kubectl -n "${APPLICATION}" create secret generic mkcert --from-file=rootCA.pem="$(mkcert -CAROOT)/rootCA.pem" || echo ok
+
+if ! kubectl get configmap certifi -n "${APPLICATION}" >/dev/null 2>&1; then
+  echo "5. Inject our custom CA in a configmap for certifi"
+  curl https://raw.githubusercontent.com/certifi/python-certifi/refs/heads/master/certifi/cacert.pem -o /tmp/cacert.pem
+  cat "$(mkcert -CAROOT)/rootCA.pem" >>/tmp/cacert.pem
+  kubectl -n "${APPLICATION}" create configmap certifi --from-file=cacert.pem=/tmp/cacert.pem
+  kubectl -n "${APPLICATION}" create secret generic certifi --from-file=/tmp/cacert.pem || echo ok
+fi
+
+echo "5b. Smoke test: the ingress chain answers on https://127.0.0.1"
+# Before Tilt deploys the app this returns the styled 404 from the default
+# backend — that still proves LB -> controller works. 000 means the
+# LoadBalancer is not bound to localhost.
+HTTP_CODE=$(curl -sk -o /dev/null -w '%{http_code}' --max-time 10 https://127.0.0.1/ || true)
+if [ "${HTTP_CODE}" = "000" ]; then
+  echo "⚠️  Nothing answered on https://127.0.0.1 — check the LoadBalancer:"
+  echo "    kubectl -n ingress-nginx get svc ingress-nginx-controller"
+else
+  echo "✅ Ingress reachable (HTTP ${HTTP_CODE})"
+fi
+
+echo "6. Check pod readiness across all namespaces..."
+
+sleep_interval=10
+
+echo "Initial wait time: $((sleep_interval * 2)) seconds…"
+sleep $((sleep_interval * 2))
+
+check_pods_ready() {
+    local max_attempts=60  # Maximum number of attempts (10 minutes with 10s intervals)
+    local attempt=1
+
+    while [ $attempt -le $max_attempts ]; do
+        echo "Attempt $attempt/$max_attempts - Checking pod status..."
+
+        not_ready_count=$( kubectl get po -A --no-headers | grep -v -E "Running|Completed"| wc -l | tr -d ' ')
+
+        if [ "$not_ready_count" -eq 0 ]; then
+            echo "✅ All pods are ready!"
+            return 0
+        else
+            echo "⏳ $not_ready_count pod(s) still not ready. Waiting $sleep_interval seconds…"
+            sleep $sleep_interval
+            ((attempt++))
+        fi
+    done
+
+    echo "❌ Timeout: Some pods are still not ready after 10 minutes"
+    echo "Final pod status:"
+    kubectl get po -A
+    return 1
+}
+
+if check_pods_ready; then
+    echo "🎉 Cluster is fully ready!"
+else
+    echo "⚠️  Some pods may need manual intervention"
+    exit 1
+fi

docs/developping_locally.md

@@ -143,3 +143,24 @@ $ make start-tilt-keycloak
 ```
 
 Monitor Tilt’s progress at [http://localhost:10350/](http://localhost:10350/). After Tilt actions finish, you can access the app at [https://meet.127.0.0.1.nip.io/](https://meet.127.0.0.1.nip.io/).
+
+### Alternative: OrbStack's built-in Kubernetes (macOS)
+
+If you use [OrbStack](https://orbstack.dev/) on macOS, you can run the stack on its built-in Kubernetes cluster instead of kind. It uses noticeably less RAM (no nested kubeadm node container) and no local registry is needed: OrbStack's cluster shares the Docker image store, so Tilt uses images directly without pushing.
+
+Enable Kubernetes in OrbStack (Settings > Kubernetes), then:
+
+```shellscript
+$ make build-k8s-cluster-orbstack
+```
+
+This installs ingress-nginx (exposed by OrbStack on `127.0.0.1:80/443`), the mkcert TLS certificates, and the CoreDNS rewrite for `*.127.0.0.1.nip.io`, then you start Tilt as usual:
+
+```shellscript
+$ make start-tilt-keycloak
+```
+
+Notes:
+- Ports 80/443 must be free: delete the kind cluster first if you used it (`kind delete cluster --name suite`).
+- If you "Reset Kubernetes" in OrbStack, re-run `make build-k8s-cluster-orbstack`.
+- kind remains the reference setup (matches CI and lets you pin the Kubernetes version).