Skip to content

Docker Swarm

The smallest orchestrator that gives you restart policies, placement constraints, and a controlled update order.

docker swarm init
docker node update --label-add multica.runtime=true "$(hostname)"

set -a; . ./.env; set +a
docker stack deploy -c deploy/swarm/stack.yml multica-node

docker stack deploy ignores env_file, so the stack interpolates from the shell — export .env first.

One stack per role

MULTICA_IMAGE=ghcr.io/proreka/multica-runtime:node \
MULTICA_DAEMON_ID=multica-node \
  docker stack deploy -c deploy/swarm/stack.yml multica-node

MULTICA_IMAGE=ghcr.io/proreka/multica-runtime:php \
MULTICA_DAEMON_ID=multica-php \
  docker stack deploy -c deploy/swarm/stack.yml multica-php

Each role has its own stack, service, volume and daemon id.

Why stop-first

The stack uses:

update_config:
  order: stop-first

A rolling update would briefly run two containers with the same MULTICA_DAEMON_ID. They would compete for tasks. Stop-first makes the old container deregister before the new one starts.

Autoscaler

set -a; . ./.env; set +a
docker stack deploy -c deploy/swarm/autoscaler.yml multica-autoscaler

It must run on a manager and needs the Docker socket. See Autoscaling for the configuration.

Logs

docker service logs -f multica-node_runtime
docker service ps multica-node_runtime