Docker — Compose reference¶
Core instance (Tier 1)¶
One service per host; DB, Vault and the load balancer are external to this stack.
services:
lyndrix:
image: ghcr.io/lyndrix-platform/lyndrix-core:1.4.0
restart: unless-stopped
ports: ["8081:8081"] # only reachable by the load balancer
environment:
ENV_TYPE: prod
LYNDRIX_CLUSTER_MODE: cluster
LYNDRIX_INSTANCE_ID: core-a # unique per host
LYNDRIX_INSTANCE_ROLES: api,leader # add ",runner" on the host with the Docker socket
LYNDRIX_UI_ENGINE: react
LYNDRIX_TRUSTED_PROXIES: 10.0.0.10 # the LB / VIP
LYNDRIX_DRAIN_SECONDS: "10"
DB_HOST: 10.0.0.20 # writer endpoint (MaxScale) or the single MariaDB
DB_NAME: lyndrix_db
DB_USER: lyndrix
DB_PASSWORD: ${DB_PASSWORD}
VAULT_URL: http://10.0.0.30:8200
VAULT_NODES: http://10.0.0.31:8200,http://10.0.0.32:8200,http://10.0.0.33:8200 # Tier 2
LYNDRIX_MASTER_KEY: ${LYNDRIX_MASTER_KEY}
STORAGE_SECRET: ${STORAGE_SECRET} # identical on every host
LYNDRIX_ADMIN_PASSWORD: ${LYNDRIX_ADMIN_PASSWORD}
LYNDRIX_BOT_PASSWORD: ${LYNDRIX_BOT_PASSWORD}
volumes:
- /srv/lyndrix/security:/data/security # vault_keys.enc — same file on every host
- /srv/lyndrix/storage:/data/storage # shared mount (NFS/CIFS) or local, see storage page
- /srv/lyndrix/plugins:/app/plugins # local; instances reconcile from the DB
healthcheck:
test: ["CMD", "python3", "-c", "import urllib.request,sys; sys.exit(0 if urllib.request.urlopen('http://127.0.0.1:8081/api/ready', timeout=3).status == 200 else 1)"]
interval: 10s
timeout: 5s
retries: 3
start_period: 90s
stop_grace_period: 30s # ≥ LYNDRIX_DRAIN_SECONDS + uvicorn graceful shutdown
Runner host: add - /var/run/docker.sock:/var/run/docker.sock and the HOST_* paths the IaC
Orchestrator needs, and LYNDRIX_INSTANCE_ROLES: api,leader,runner.
Frontend¶
lyndrix-ui:
image: ghcr.io/lyndrix-platform/lyndrix-ui:1.4.0
environment:
LYNDRIX_BACKEND_URL: http://10.0.0.10 # the LB / VIP, never one instance
The dev test bench¶
docker/docker-compose.ha.dev.yml is an overlay for the dev stack:
docker compose -f docker/docker-compose.dev.yml -f docker/docker-compose.ha.dev.yml up -d
curl http://localhost:8090/api/live # alternates dev-a / dev-b
curl http://localhost:8081/api/ready # instance A directly
curl http://localhost:8082/api/ready # instance B directly
It gives both cores LYNDRIX_CLUSTER_MODE=cluster, the db coordinator, LYNDRIX_DRAIN_SECONDS=5
and the same plugin, storage and security mounts, and adds the Traefik entrypoint ha (:8090)
with the health check on /api/ready and flushInterval: -1 for SSE. Instance A keeps the
runner role (IaC Orchestrator), B is api,leader and shows the orchestrator as remote.
Point the React dev shell at http://traefik:8090 (container network) to drive the whole UI
through the balancer.
Notes¶
stop_grace_periodmust exceed the drain plus uvicorn's--timeout-graceful-shutdown, otherwise Docker kills the container mid-drain.start_periodcovers the boot (Vault → DB → IAM → modules);/api/readyis 503 until then.- Both dev instances write
/app/logs/lyndrix.loginto the same mounted directory — harmless for development, never do that in production (use stdout).