Merge pull request 'fix(ci): auto-discover reachable Portainer URL from job container' (#5) from fix/runner-portainer-network into main

.gitea/workflows/ci-cd.yaml

@@ -62,53 +62,102 @@ jobs:
       #   PORTAINER_ENDPOINT_ID  Numeric endpoint ID (usually 1)
       #   PORTAINER_STACK_ID     Numeric stack ID to delete before recreating
       # ----------------------------------------------------------------------
-      - name: Redeploy stack on Portainer via SSH
+      - name: Redeploy stack on Portainer
         if: github.event_name == 'workflow_dispatch' && inputs.deploy == true
         env:
-          NAS_SSH_KEY: ${{ secrets.NAS_SSH_KEY }}
-          NAS_HOST: ${{ secrets.NAS_HOST }}
-          NAS_USER: ${{ secrets.NAS_USER }}
-          STACK_NAME: davidaragon-portfolio
-          COMPOSE_FILE: docker-compose.prod.yml
+          PORTAINER_TOKEN: ${{ secrets.PORTAINER_TOKEN }}
+          PORTAINER_ENDPOINT_ID: ${{ secrets.PORTAINER_ENDPOINT_ID }}
+          PORTAINER_STACK_ID: ${{ secrets.PORTAINER_STACK_ID }}
         run: |
-          # The act_runner job container can't reach 'portainer' because it
-          # doesn't share the runner's network namespace in v0.6.1. We work
-          # around this by SSHing into the NAS (where Portainer is) and
-          # using the docker CLI directly to do `docker stack deploy`.
           set -euo pipefail
 
-          mkdir -p ~/.ssh
-          echo "$NAS_SSH_KEY" > ~/.ssh/id_ed25519
-          chmod 600 ~/.ssh/id_ed25519
+          echo "--- Step 0: auto-discover a reachable Portainer URL ---"
+          # The runner creates an ephemeral docker network per job, so the
+          # canonical PORTAINER_URL (e.g. http://192.168.1.30:9000) often isn't
+          # reachable from inside the job container. Strategy: probe every IP
+          # we can find (container's own IPs + default gateways) against
+          # :9000/api/status and use the first one that responds 200.
+          PORTAINER_URL=""
+          CANDIDATES=()
 
-          echo "--- Pre-flight: ensure NAS is reachable ---"
-          # NAS_HOST can be either "host" or "host:port"
-          NAS_SSH_PORT=$(echo "$NAS_HOST" | grep -q ':' && echo "${NAS_HOST##*:}" || echo "22")
-          NAS_SSH_HOST="${NAS_HOST%%:*}"
-          echo "Using NAS_SSH_HOST=$NAS_SSH_HOST NAS_SSH_PORT=$NAS_SSH_PORT"
-          ssh -o StrictHostKeyChecking=no -o ConnectTimeout=5 -p "$NAS_SSH_PORT" \
-              "$NAS_USER@$NAS_SSH_HOST" \
-              'export PATH=/share/CACHEDEV1_DATA/.qpkg/container-station/usr/bin:$PATH; hostname && docker version --format "{{.Server.Version}}"' \
-              | head
+          # 1) Container's own IPv4 addresses
+          for ip in $(hostname -I 2>/dev/null | tr ' ' '\n' | grep -E '^[0-9]+\.'); do
+            CANDIDATES+=("http://${ip}:9000")
+          done
 
-          echo "--- Reading docker-compose.prod.yml from the repo ---"
-          if [ ! -f "$COMPOSE_FILE" ]; then
-            echo "ERROR: $COMPOSE_FILE not found in repo root" >&2
+          # 2) Default gateways of every default route
+          while IFS= read -r gw; do
+            [ -n "$gw" ] && CANDIDATES+=("http://${gw}:9000")
+          done < <(ip -4 route show default 2>/dev/null | awk '{print $3}' | sort -u)
+
+          # 3) Fallback: secret value (in case everything else fails)
+          CANDIDATES+=("${{ secrets.PORTAINER_URL }}")
+
+          echo "Candidates: ${CANDIDATES[@]}"
+          for url in "${CANDIDATES[@]}"; do
+            code=$(curl -sS -o /dev/null -w '%{http_code}' --max-time 3 "${url}/api/status" 2>/dev/null || echo "000")
+            echo "  probe ${url}/api/status -> ${code}"
+            if [ "$code" = "200" ]; then
+              PORTAINER_URL="$url"
+              echo "  -> using ${PORTAINER_URL}"
+              break
+            fi
+          done
+
+          if [ -z "$PORTAINER_URL" ]; then
+            echo "ERROR: no candidate URL reached Portainer. Tried: ${CANDIDATES[@]}" >&2
             exit 1
           fi
 
-          echo "--- Redeploying stack '$STACK_NAME' on the NAS ---"
-          # The NAS has Portainer and docker compose (v2) available. We stream
-          # the compose file over SSH and let docker compose recreate the
-          # project. We pin the project name to 'davidaragon-portfolio' so the
-          # volumes and networks of the existing stack are reused.
-          cat "$COMPOSE_FILE" | ssh -o StrictHostKeyChecking=no -p "$NAS_SSH_PORT" \
-              "$NAS_USER@$NAS_SSH_HOST" \
-              "export PATH=/share/CACHEDEV1_DATA/.qpkg/container-station/usr/bin:\$PATH; cd /tmp && docker compose -p '$STACK_NAME' -f - up -d"
+          echo "--- Step 1: pre-flight (delete existing stack if present) ---"
+          DELETE_HTTP_CODE=$(curl -sS -o /tmp/portainer-delete.json -w '%{http_code}' \
+            -X DELETE \
+            -H "X-API-Key: ${PORT...EN}" \
+            "${PORTAINER_URL}/api/stacks/${PORTAINER_STACK_ID}?endpointId=${PORTAINER_ENDPOINT_ID}")
+          echo "DELETE HTTP ${DELETE_HTTP_CODE}"
+          if [ "${DELETE_HTTP_CODE}" != "204" ] && [ "${DELETE_HTTP_CODE}" != "404" ]; then
+            echo "ERROR: Portainer rejected DELETE on stack ${PORTAINER_STACK_ID}:" >&2
+            cat /tmp/portainer-delete.json >&2
+            exit 1
+          fi
 
-          echo "--- Stack '$STACK_NAME' redeployed. Smoke-test: ---"
-          sleep 5
-          curl -sS -o /dev/null -w 'http://localhost:3001/ -> HTTP=%{http_code}\n' \
-              --max-time 5 http://localhost:3001/ || true
+          echo "--- Step 2: read docker-compose.prod.yml ---"
+          if [ ! -f docker-compose.prod.yml ]; then
+            echo "ERROR: docker-compose.prod.yml is missing from the repo root" >&2
+            exit 1
+          fi
+          # Inline the compose file. Portainer expects `composeFileContent` as raw text.
+          COMPOSE_BODY=$(jq -Rs --arg compose "$(cat docker-compose.prod.yml)" \
+            '{composeFileContent: $compose, env: []}' < /dev/null)
 
-          echo "--- Done ---"
+          echo "--- Step 3: create fresh stack from docker-compose.prod.yml ---"
+          CREATE_HTTP_CODE=$(curl -sS -o /tmp/portainer-create.json -w '%{http_code}' \
+            -X POST \
+            -H "X-API-Key: ${PORTAINER_TOKEN}" \
+            -H "Content-Type: application/json" \
+            --data "${COMPOSE_BODY}" \
+            "${PORTAINER_URL}/api/stacks?endpointId=${PORTAINER_ENDPOINT_ID}&type=2&method=string&name=davidaragon-portfolio")
+          echo "CREATE HTTP ${CREATE_HTTP_CODE}"
+          if [ "${CREATE_HTTP_CODE}" != 201 ]; then
+            echo "ERROR: Portainer rejected stack creation:" >&2
+            cat /tmp/portainer-create.json >&2
+            exit 1
+          fi
+
+          echo "--- Step 4: smoke-test the freshly deployed stack ---"
+          # Give the container a brief window to start before checking.
+          sleep 8
+          # The Portainer host:9000 is reachable from the job container
+          # (we just proved that with the auto-discovery step above). The
+          # portfolio container itself is published on host port 3001, so
+          # smoke-test through the same host. We don't fail the job if the
+          # proxy upstream isn't reachable from the runner's network.
+          HEALTH=$(curl -sS -o /dev/null -w '%{http_code}' \
+            --max-time 5 \
+            "${PORTAINER_URL%:[0-9]*}:3001/" 2>/dev/null || true)
+          echo "Health check on ${PORTAINER_URL%:[0-9]*}:3001/ returned: ${HEALTH:-<timeout/unreachable>}"
+          # We log but don't fail the job if 3001 isn't reachable — the upstream
+          # proxy (Nginx Proxy Manager → davidaragon.impresion3d.pro) is a better
+          # place to wire a hard-fail check in a future iteration.
+
+          echo "--- Stack recreated successfully. ---"