Nothing alerts when ipx goes wrong #365
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Found reviewing what production's traces and metrics leave out, 2026-10-25. Prometheus has no rules and Grafana no alerting provisioned, so a stuck worker, a full pool or ipx down waits for someone to run the prod check; Tempo's search was broken for 15 hours on 2026-10-25 unnoticed. Wanted: Grafana alert rules from the repo, beside the dashboard: ipx or the directory down, the worker stuck (last scan over 15 min, a job over 30 min), the pool full, 5xx, Jev rate-limited or malformed, Tower's load and memory, Tempo's block poll failing. No contact point yet: they show as firing in Grafana's Alerting page until one is added.
Fixed in
dd91602and 34-line follow-upc4f0a36: grafana/alerts.py writes 16 Grafana alert rules to the monitoring project's grafana-provisioning/alerting/ipx.json (ipx, directory or ipx-web down; scans stopped; a job over 30 min; the pool full; 5xx; 502s; Jev refused; Signals stopped; the directory's refresher; Tower's load and memory; monitoring down; Tempo's blocks; iPX's lines not reaching Loki). No contact point yet: firing shows in Grafana's Alerting page. All healthy after the deploy. Deployed 2026-10-25.