swarm-production

Production Docker Swarm Infrastructure

Cluster Overview

Nodes

• p0 (Manager/Leader) - Infrastructure services
• p1 (Manager) - Application services
• p2 (Manager) - Application services
• p3 (Manager) - Application services

All nodes are managers providing a 4-node quorum (can tolerate 2 node failures while maintaining quorum).

Storage

• GlusterFS mounted at /home/doc/swarm-data/ on all nodes
• Shared storage enables services to run on any node without storage constraints

Service Distribution Strategy

Pinned Services

Services that must run on specific nodes:

• traefik (p0) - Published ports 80/443, needs stable IP for DNS
• portainer (p0) - Management UI, stays with leader for convenience
• rsync (manager constraint) - Backup service, needs manager access

Floating Services

Services that can run on any node (swarm auto-balances):

• adminer
• authentik (server, worker, redis)
• n8n
• paperless (webserver, redis)
• tracker-nginx
• uptime-kuma

Recent Changes (2025-10-30)

Swarm Rebalancing

• Promoted p1, p2, p3 from workers to managers
• Removed unnecessary hostname constraints from service configs
• Force-redeployed services to redistribute across all nodes
• Verified GlusterFS accessibility on all nodes

Results

• Achieved balanced workload distribution across all 4 nodes
• Improved high availability with 4-node manager quorum
• Services now self-balance automatically when nodes fail/recover
• Fixed Portainer agent connectivity by restarting agents after manager promotion