fix(ci): deploy via SSH to NAS instead of Portainer API

act_runner v0.6.1 does not propagate the runner's network namespace
to job containers, so the job can never resolve 'portainer' or reach
'192.168.1.30:9000'. After trying every combination of
container.network_mode / container.network / options in the runner
config, the only reliable solution is to skip the in-cluster Portainer
API call and do the deploy via SSH directly on the NAS.

- New secrets: NAS_SSH_KEY, NAS_HOST, NAS_USER.
- The new 'Redeploy stack on Portainer via SSH' step writes the SSH key
  to ~/.ssh, then pipes docker-compose.prod.yml to
  'docker compose -p davidaragon-portfolio -f - up -d' over SSH.
- The old Portainer-API step is removed.

Smoke-test at the end hits the public port on the host (3001) and
logs the HTTP code; we don't fail the job if the public proxy
upstream isn't reachable from the runner's network.
This commit is contained in:
2026-07-10 12:37:16 +00:00
parent 89150ed41f
commit 6ff7bc5fda
+37 -62
View File
@@ -62,73 +62,48 @@ jobs:
# PORTAINER_ENDPOINT_ID Numeric endpoint ID (usually 1)
# PORTAINER_STACK_ID Numeric stack ID to delete before recreating
# ----------------------------------------------------------------------
- name: Redeploy stack on Portainer
- name: Redeploy stack on Portainer via SSH
if: github.event_name == 'workflow_dispatch' && inputs.deploy == true
env:
PORTAINER_TOKEN: ${{ secrets.PORTAINER_TOKEN }}
PORTAINER_ENDPOINT_ID: ${{ secrets.PORTAINER_ENDPOINT_ID }}
PORTAINER_STACK_ID: ${{ secrets.PORTAINER_STACK_ID }}
NAS_SSH_KEY: ${{ secrets.NAS_SSH_KEY }}
NAS_HOST: ${{ secrets.NAS_HOST }}
NAS_USER: ${{ secrets.NAS_USER }}
STACK_NAME: davidaragon-portfolio
COMPOSE_FILE: docker-compose.prod.yml
run: |
# Portainer, Gitea and the runners all share the `portainer_default`
# Docker network on the NAS, so the service name "portainer" resolves
# directly from the job container on port 9000.
PORTAINER_URL="http://portainer:9000"
echo "Using PORTAINER_URL=$PORTAINER_URL"
if ! curl -sS -o /dev/null -w '%{http_code}' --max-time 5 \
"${PORTAINER_URL}/api/status" | grep -q '^200$'; then
echo "ERROR: Portainer not reachable at $PORTAINER_URL" >&2
# The act_runner job container can't reach 'portainer' because it
# doesn't share the runner's network namespace in v0.6.1. We work
# around this by SSHing into the NAS (where Portainer is) and
# using the docker CLI directly to do `docker stack deploy`.
set -euo pipefail
mkdir -p ~/.ssh
echo "$NAS_SSH_KEY" > ~/.ssh/id_ed25519
chmod 600 ~/.ssh/id_ed25519
echo "--- Pre-flight: ensure NAS is reachable ---"
ssh -o StrictHostKeyChecking=no -o ConnectTimeout=5 \
"$NAS_USER@$NAS_HOST" 'hostname && docker version --format "{{.Server.Version}}"' \
| head
echo "--- Reading docker-compose.prod.yml from the repo ---"
if [ ! -f "$COMPOSE_FILE" ]; then
echo "ERROR: $COMPOSE_FILE not found in repo root" >&2
exit 1
fi
echo "--- Step 1: pre-flight (delete existing stack if present) ---"
DELETE_HTTP_CODE=$(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 ${PORTAINER_STACK_ID}:" >&2
cat /tmp/portainer-delete.json >&2
exit 1
fi
echo "--- Redeploying stack '$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 "$COMPOSE_FILE" | ssh -o StrictHostKeyChecking=no \
"$NAS_USER@$NAS_HOST" \
"export PATH=/share/CACHEDEV1_DATA/.qpkg/container-station/usr/bin:\$PATH; cd /tmp && docker compose -p '$STACK_NAME' -f - up -d"
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: $compose, env: []}' < /dev/null)
echo "--- Stack '$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 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. ---"
echo "--- Done ---"