Task · TSO-0205

Alert on a stale persistent flow-store checkpoint

Description

tailscale2otel.flow_store.last_checkpoint_timestamp (internal/appcatalog/catalog.go) has no rule; a stuck SQLite WAL checkpoint grows the journal unbounded. Only deployments with the persistent backend emit it.

Acceptance Criteria

Definition of Done

Implementation Plan

Loop33 checkpoint lane: generated persistent-backend stale checkpoint alert with stale/fresh/absent promtool fixtures.

Implementation Notes

loop33 landed4c84259bf43a5295a69731caf8d25446877454df in tipfdaa4d8b84c44cde5ca99c00f6f4256886a8b409. Exact composed gate20261101T212748Z-tailscale2otel-9e155cc2 exit0 including regeneration/no-diff. Real stale/fresh/absent/recovery fixtures and independent guarded PASS. Default hourly checkpoint compared with2h threshold held15m; longer custom sweep intervals require documented tuning. Live publication separate.

Final Summary

Shipped persistent-store checkpoint-staleness alert with explicit threshold and absent-series suppression.

References

View the source file on GitHub