Forgejo proxy watchdog fails a task that is still queued in Solstice (STALE_JOB_TIMEOUT counts queue time) #10
Loading…
Reference in a new issue
No description provided.
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?
Observed 2026-09-20 right after the executor lane was unwedged (see #9):
solstice.jobs.v1had ~60 ready messages, so a newly pushed Anima commit (25aa89e0, run 8787) waited in the Solstice queue for more than 30 minutes. Its Forgejotestjob was then marked failed withwhile the Solstice job for the same commit was still
queuedand later ran fine. The proxy's stale-task watchdog starts counting when the JobRequest is published, so any queue backlog longer than STALE_JOB_TIMEOUT_SECS (1800) turns green work into red Forgejo runs, and the only remedy is a new push.Expected: the watchdog should count from when the orchestrator actually starts the job (or reset when the job is observed
queued/runningin Solstice), and ideally the proxy should keep the Forgejo task alive with an UpdateTask/log line ("queued behind N jobs") while waiting.