Plex exporter¶

Purpose¶

kubernetes/apps/observability/plex-exporter runs timothystewart6/prometheus-plex-exporter against the in-cluster Plex server and alerts on scrape health, server reachability and transcode-driven host CPU.

Why a high restart count is expected¶

The exporter treats a failed Plex websocket read as fatal. It logs "cannot listen to plex server events" and exits instead of reconnecting itself, so the pod's Kubernetes restart is its reconnect path. It has no backoff for a Plex server that is not there yet, so while Plex is down, during a Plex image update for example, the pod hot-loops and its restart count jumps in bursts of 5 to 12 an hour, then sits still for days once Plex is back.

Over the seven days to 2026-07-26 that cost about 25 minutes of scrape gaps, 99.75% availability, with no restarts after 2026-07-21 (#3845). A restart-count alert was considered and dropped for two reasons. The bursts already trip KubePodCrashLooping, the chart's default crash-loop rule, and PlexServerUnreachable covers the case where the exporter stays up but cannot reach Plex. This workload's app-template container key is app, not plex-exporter, so a rule written against container="plex-exporter" would match nothing and silently never fire.