A hub restart can end a running Flow job as an error issue #111 closed: completed andymccall opened this · 1 comment
andymccall
Correction after reading the job's error: it was not the restart. earth-1's workspace fetch failed with "Temporary failure in name resolution" for oxygenhub.io, a transient DNS failure, and the job ended at once. The fix is in the runner: retry the fetch (and subrepo fetches) on transient network errors (DNS, connect, timeouts, 502/503/504) with backoff for about a minute before reporting an error.
andymccall closed ·
Run issue #144 (2026-10-04): the test job on earth-1 ended as
errorafter 10 s with no log, at the moment the hub restarted for a deploy. A runner whose log or finish call fails while the hub restarts should retry with backoff for a minute before giving up, and the hub should not end a job as an error because a call came while it was down. Reran as issue #145.