Rolling Restarts

Restart rninja-cached without disrupting builds.

Single Server

# Graceful restart
systemctl restart rninja-cached

Clients will retry automatically.

With Load Balancer

1. Remove server from pool
2. Wait for active requests to complete
3. Restart server
4. Add back to pool

# Remove from LB
curl -X POST http://lb/servers/cache1/disable

# Wait and restart
sleep 30
systemctl restart rninja-cached

# Re-enable
curl -X POST http://lb/servers/cache1/enable

Client Resilience

Clients handle restarts gracefully:

• Auto-retry on connection failure
• Fall back to local cache (mode=auto)
• Reconnect to new server

Best Practices

1. Use auto mode for client resilience
2. Set reasonable timeouts
3. Monitor during restart
4. Restart during low-usage periods