Session Affinity Loss After Cache Node Failover

A session cache node has failed over. Request latency and error rates look almost normal — the cache cluster isn't down and isn't slow. But a subset of logged-in users are being bounced between inconsistent session states: some see themselves abruptly logged out, others see a stale cart. The load balancer's consistent-hash ring is still routing against the pre-failover cache topology, sending some requests to a node that never received that session's replicated data. Diagnose the routing/replication mismatch and restore session consistency without just throwing capacity at a cluster that was never actually overloaded.

Difficulty

Staff

Expected Time

~15 min

Category

Caching