fix(bug-hunter): round 6 - cluster health/learner gate, preauth timeout, route rename, sink cleanup, CI exit codes

Component view — Cluster Lifecycle & Mesh

Component view — RTMP Routing & Media Outputs

Clearing stale routes across shards on publish rename

fix(bug-hunter): round 6 - cluster health/learner gate, preauth timeout, route rename, sink cleanup, CI exit codesOpenRTMP/librtmp2-server #273 · 5 diagrams

◈ Rendered by PR Lens · crafted with ❤️ by the Coldtea team

100%

01 / 06Shard now ends previous route on stream renameWhen a connection switches stream keys, other shards drop their old claim right away instead of waiting 120 seconds. records connection routes in +28 −5 and sends .
02 / 06Resumed learner now stays publish-gatedA restarting node checks its voter status and stays in learner state until promoted to voter. checks membership in +17 −1 before sets node state.
03 / 06Peer handshake now times out on stalled helloMedia peer connections abort after 5 seconds if the peer authenticates but stalls before sending hello. adds a timeout for hello frames in +3 −1.
04 / 06Output shutdown now marks failed workerIf a sink worker stalls for 5 seconds during stop, it flags failure so the monitor kills the child process. sets failed flag during shutdown in +4 −0.
05 / 06Cluster manager now drops removed node healthRemoving a node from the cluster now deletes its health record alongside its peer and network state. deletes health state in +17 −1 on node removal.
06 / 06Everything this change touchedRoute eviction, learner health gating, peer handshake timeouts, and worker process cleanup in one view.

esc to leave