bvandeusen
6cf70e22db
Mitigation for the nightly fabledscribe Postgres outage on the vdnt-docker02 Swarm node (incidents 2026-05-15/16/17 around 03:50 UTC). Confirmed kill chain (not the trigger): a brief host-level setns/exec stall makes the Docker healthcheck exec fail with exit 1 → unhealthy → SIGKILL → fast-shutdown can't finish on NFS in 10s → exit 137 → swarm restart_policy.max_attempts: 5 burns out → DB stays dead. Hardens the `db` service so a transient host blip can't escalate to killing the database: - stop_grace_period: 120s (gives PG room to fsync on shutdown) - healthcheck: interval 30s / timeout 10s / retries 10 / start_period 180s (only gates app startup order — not authoritative liveness) - prod: restart_policy condition=on-failure, max_attempts=0, window=120s - quickstart/dev: restart: unless-stopped Host-side trigger (what stalls runc/exec at ~03:50 UTC) is still under investigation — see project_pg_nightly_outage.md. Note: the Portainer prod stack differs from docker-compose.prod.yml here (NFS bind, traefik labels, no ollama). The same `db:` block needs to be pasted into Portainer for the prod mitigation to apply. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
76 lines
2.6 KiB
YAML
76 lines
2.6 KiB
YAML
services:
|
|
app:
|
|
build: .
|
|
ports:
|
|
- "5000:5000"
|
|
depends_on:
|
|
db:
|
|
condition: service_healthy
|
|
ollama:
|
|
condition: service_started
|
|
volumes:
|
|
- app_data:/data
|
|
# To use a bind mount instead (gives direct host access to all app data):
|
|
# - ./data:/data
|
|
environment:
|
|
DATABASE_URL: "postgresql+asyncpg://${POSTGRES_USER:-fabled}:${POSTGRES_PASSWORD:-fabled}@db:5432/${POSTGRES_DB:-fabledassistant}"
|
|
OLLAMA_URL: "http://ollama:11434"
|
|
OLLAMA_MODEL: "${OLLAMA_MODEL:-qwen3:8B}"
|
|
SECRET_KEY: "${SECRET_KEY:-dev-secret-change-me}"
|
|
# Uncomment and set to enable web research and image search via SearXNG:
|
|
# SEARXNG_URL: "http://searxng:8080"
|
|
# IMAGE_CACHE_DIR: /data/images # default, change if using a different mount path
|
|
# IMAGE_MAX_BYTES: "5242880" # 5 MB per image, adjust if needed
|
|
# Push notifications (VAPID keys - generate with: python -c "from py_vapid import Vapid01; v=Vapid01(); v.generate_keys(); print(v.private_key, v.public_key)")
|
|
VAPID_PRIVATE_KEY: "${VAPID_PRIVATE_KEY:-}"
|
|
VAPID_PUBLIC_KEY: "${VAPID_PUBLIC_KEY:-}"
|
|
VAPID_CLAIMS_SUB: "${VAPID_CLAIMS_SUB:-mailto:admin@fabledassistant.local}"
|
|
healthcheck:
|
|
test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:5000/api/health')"]
|
|
interval: 10s
|
|
timeout: 5s
|
|
retries: 5
|
|
start_period: 30s
|
|
|
|
db:
|
|
image: postgres:16-alpine
|
|
stop_grace_period: 120s
|
|
restart: unless-stopped
|
|
volumes:
|
|
- pgdata:/var/lib/postgresql/data
|
|
environment:
|
|
POSTGRES_USER: ${POSTGRES_USER:-fabled}
|
|
POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-fabled}
|
|
POSTGRES_DB: ${POSTGRES_DB:-fabledassistant}
|
|
# Lenient by design: a transient host exec/healthcheck stall must never
|
|
# escalate to killing the DB. Health here only gates app startup order.
|
|
healthcheck:
|
|
test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER:-fabled}"]
|
|
interval: 30s
|
|
timeout: 10s
|
|
retries: 10
|
|
start_period: 180s
|
|
|
|
ollama:
|
|
image: ollama/ollama
|
|
volumes:
|
|
- ollama_models:/root/.ollama
|
|
environment:
|
|
OLLAMA_MAX_LOADED_MODELS: "2"
|
|
OLLAMA_NUM_PARALLEL: "2"
|
|
OLLAMA_KEEP_ALIVE: "30m"
|
|
OLLAMA_FLASH_ATTENTION: "1"
|
|
# GPU reservation commented out — no nvidia-container-toolkit on this host
|
|
# deploy:
|
|
# resources:
|
|
# reservations:
|
|
# devices:
|
|
# - driver: nvidia
|
|
# count: all
|
|
# capabilities: [gpu]
|
|
|
|
volumes:
|
|
pgdata:
|
|
ollama_models:
|
|
app_data:
|