fleetd #512 follow-up: record why the drain-complete line must not move into a finally #547

Merged
ltms merged 1 commits from fix/512-drain-line-non-goal into main 2026-09-12 09:29:47 +02:00

1 Commits

Author SHA1 Message Date
Dai Ha 5eb4267a4a fleetd #512 follow-up: record why the drain-complete line must not move into a finally
CI / contract (pull_request) Successful in 1m3s
CI / build (pull_request) Successful in 2m8s
The javadoc said the log.info fires "every time", which reads as an invitation
to the exact edit that destroys it. The absence of the line is the signal that
the drain died, so a finally would remove the signal and print partial counts in
the same change.

Raised by the fleet01 lead from their 2026-09-10 incident: that drain is known to
have died only because it threw and left a stack trace. A drain that hung, or
returned early on a condition, leaves no trace, no ERROR token and no priority --
only a missing line.

Comment only. No behaviour change.
2026-09-12 14:21:15 +07:00