Multi-site failover

MinIO Sidekick supports automatic failover between multiple sites when backend servers become unavailable.

Site configuration

Configure multiple sites by passing separate groups of backend servers as command arguments. Sidekick tries sites in the order specified, routing traffic to the first site with healthy backends. When all backends in the active site fail health checks, the load balancer automatically switches to the next site.

Each site can contain multiple backend servers. Sidekick distributes requests across healthy backends within the active site using the configured load balancing algorithm. Only when all backends in a site become unhealthy does failover occur to the next site.

Failover behavior

Site failover happens automatically when health checks detect that all backends in the current site are down. Sidekick immediately begins routing requests to the next configured site. No manual intervention or DNS changes are required.

Sidekick picks a site for every request, and always uses the first healthy site in the order specified. When backends in an earlier site recover, Sidekick fails back to that site automatically.

Read-only failover

For multi-site deployments with read-write splitting, configure a read-only health check path for the last site using --read-health-path. Sidekick then treats the last site as healthy when it meets read quorum, even without write quorum. Sidekick does not separate reads from writes. After failover, it forwards every request to that site, including writes. Use your clients or another routing layer to stop writes to a read-only failover site.