Don't trace commands that never finish #86
Loading…
Reference in a new issue
No description provided.
Delete branch "patrick/no-trace-for-daemons"
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?
Dallas's v0.3.30 upgrade timed out because Horizon crash-looped:
Memory limit exceeded: Using 65/64MB, exit 12, with OpenTelemetry reportingmissing call to Scope::detach() for scope #3096.Cause: command tracing opens a root span on
CommandStartingand closes it onCommandFinished.horizon(and thehorizon:supervisor/horizon:workprocesses it starts),queue:work,reverb:startandschedule:worknever finish, so their span stayed open for the life of the pod. Every job and query nested under it, and nothing was ever exported. Tracing only became effective on Dallas once its endpoint pointed at the real Tempo, which is why this didn't show before.Fix: those commands open no span. Each job a worker runs now gets its own trace through the queue seam, which is what job tracing intended.
Checked: 3 new tests in TracingTest; the two daemon ones fail against the old code. Full suite: 516 passed, 8 skipped. Pint clean.
🤖 Generated with Claude Code