Cache regions now run a warm standby replica on the same box with automatic failover. During a rolling deploy the controller fails the cache Service over to a Ready standby that has already replicated the latest state via the async outbox, so writes and serving no longer stall while a primary rolls. Only a brief window of misses on the very newest artefacts is possible, and those regenerate naturally on the next build.
Hive
Gapless Kura cache deploys with no write stall
Published
Jul 01, 2026 · 07:46 UTC
Repository
tuist/tuist