diff --git a/infrastructure/docker-compose.gpu-box.yml b/infrastructure/docker-compose.gpu-box.yml index a06e45c97..1c8185108 100644 --- a/infrastructure/docker-compose.gpu-box.yml +++ b/infrastructure/docker-compose.gpu-box.yml @@ -299,7 +299,12 @@ services: depends_on: loki: condition: service_started - # healthcheck disabled: promtail image has no curl/wget/nc; restart policy handles crashes + healthcheck: + test: ['CMD', 'bash', '-c', 'exec 3<>/dev/tcp/loki/3100'] + interval: 60s + timeout: 5s + retries: 3 + start_period: 15s # ============================================ # Phase 2d — Glitchtip mit dedizierter Postgres + Redis (2026-05-06) @@ -378,6 +383,12 @@ services: SECRET_KEY: ${GLITCHTIP_SECRET_KEY} GLITCHTIP_DOMAIN: https://glitchtip.mana.how CELERY_WORKER_AUTOSCALE: '1,3' + healthcheck: + test: ['CMD', 'bash', '-c', 'exec 3<>/dev/tcp/glitchtip-redis/6379'] + interval: 60s + timeout: 5s + retries: 3 + start_period: 30s # ============================================ # Phase 2e — Status-Page (2026-05-07): generator + nginx auf GPU-Box. @@ -412,6 +423,12 @@ services: sh /tmp/generate.sh sleep 60 done + healthcheck: + test: ['CMD-SHELL', '[ -f /output/status.json ] && [ $$(( $$(date +%s) - $$(stat -c %Y /output/status.json) )) -lt 180 ]'] + interval: 90s + timeout: 5s + retries: 2 + start_period: 60s status-nginx: image: nginx:alpine