failover

Switching work from a failed or unavailable system to a standby copy so the service keeps running.

In more detail

Failover can be automatic, driven by health checks, or started by people, and it usually involves pointing users at the standby, for example by changing a DNS record. Failback is the return trip once the original system is fixed. Failover protects availability, not data: a standby database still contains any bad data the primary had.

Examples

"hot": {"replica already current": 0, "switch DNS and test": 0.5}
Probes and upstream timeouts on the Ingress let it stop sending users to dead pods.

Related terms

Taught in

References

Certification courses