83 lines
2.4 KiB
YAML
83 lines
2.4 KiB
YAML
services:
|
|
operator-ui:
|
|
build: .
|
|
container_name: control-plane-ui
|
|
user: "1000:1000"
|
|
command: python src/operator_ui.py
|
|
ports:
|
|
- "18180:8080"
|
|
volumes:
|
|
- /opt/homelab:/opt/homelab
|
|
restart: unless-stopped
|
|
healthcheck:
|
|
test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8080/', timeout=3).read()"]
|
|
interval: 30s
|
|
timeout: 10s
|
|
retries: 3
|
|
|
|
observer:
|
|
build: .
|
|
container_name: control-plane-observer
|
|
user: "1000:1000"
|
|
command: python /repo/scripts/observer/observer.py
|
|
volumes:
|
|
- /opt/homelab:/opt/homelab
|
|
- ../..:/repo:ro
|
|
restart: unless-stopped
|
|
environment:
|
|
- REPO_ROOT=/repo
|
|
- RUNTIME_PATH=/opt/homelab
|
|
healthcheck:
|
|
test: ["CMD", "test", "-f", "/opt/homelab/state/observer.heartbeat"]
|
|
interval: 30s
|
|
timeout: 5s
|
|
retries: 3
|
|
start_period: 5s
|
|
|
|
supervisor:
|
|
build: .
|
|
container_name: control-plane-supervisor
|
|
user: "1000:1000"
|
|
command: python src/supervisor.py
|
|
volumes:
|
|
- /opt/homelab:/opt/homelab
|
|
- ../..:/repo:ro
|
|
restart: unless-stopped
|
|
environment:
|
|
- REPO_ROOT=/repo
|
|
- RUNTIME_PATH=/opt/homelab
|
|
healthcheck:
|
|
# Freshness, not just existence: `test -f` stays true forever once the
|
|
# heartbeat file is created once, even if the reconcile loop has been
|
|
# silently frozen for hours (see 2026-07-15 incident — container reported
|
|
# "healthy" for 24h with a dead loop). Fail if the heartbeat hasn't been
|
|
# touched in the last 180s (loop interval 30s + RECONCILE_TIMEOUT 90s +
|
|
# buffer for a legitimately slow cycle).
|
|
test: ["CMD", "python", "-c", "import os,sys,time; p='/opt/homelab/state/supervisor.heartbeat'; sys.exit(0 if os.path.exists(p) and time.time()-os.path.getmtime(p)<180 else 1)"]
|
|
interval: 60s
|
|
timeout: 5s
|
|
retries: 3
|
|
start_period: 10s
|
|
|
|
executor:
|
|
build: .
|
|
container_name: control-plane-executor
|
|
user: "1000:1000"
|
|
group_add:
|
|
- "999"
|
|
command: python src/executor.py
|
|
volumes:
|
|
- /opt/homelab:/opt/homelab
|
|
- ../..:/repo
|
|
- /var/run/docker.sock:/var/run/docker.sock
|
|
restart: unless-stopped
|
|
environment:
|
|
- REPO_ROOT=/repo
|
|
- RUNTIME_PATH=/opt/homelab
|
|
healthcheck:
|
|
test: ["CMD", "test", "-f", "/opt/homelab/state/executor.heartbeat"]
|
|
interval: 30s
|
|
timeout: 5s
|
|
retries: 3
|
|
start_period: 5s
|