docker-swarm

Best Practices

• Use version 3.8 or later for full Swarm deploy feature support (deploy, secrets, configs)
• Define deploy.replicas, update_config, rollback_config, and restart_policy for every service
• Use placement constraints to control node scheduling (node.role, node.labels, node.hostname)
• Configure health checks with test, interval, timeout, retries, and start_period for all services
• Use Docker secrets (not environment variables) for passwords, API keys, and certificates
• Use Docker configs for non-sensitive configuration files that services consume at runtime
• Define resource reservations AND limits under deploy.resources for capacity planning
• Use overlay networks with encryption enabled for secure multi-host service communication
• Set update_config.failure_action to rollback for automatic rollback on failed updates
• Pin image versions with SHA256 digests or specific tags for production reproducibility
• Use deploy.mode global for daemon-like services (logging agents, monitoring exporters)
• Configure endpoint_mode (vip or dnsrr) based on load balancing requirements
• Use deploy.labels for service metadata and integration with reverse proxies (Traefik labels)
• Set stop_grace_period for services that need time for graceful shutdown