galaxyproject / pulsar

Distributed job execution application built for Galaxy
https://pulsar.readthedocs.io
Apache License 2.0
37 stars 50 forks source link

Do not retry on 403 error #353

Closed natefoo closed 6 months ago

natefoo commented 9 months ago

When postprocessing, a 403 error typically means the job was cancelled by the user (except the strange case in #342). When this is the case, Pulsar keeps trying as per the retry settings, but the transfers will never succeed. It should probably just fail unconditionally in the event of a 403.

Alternatively, we could make configurable the error conditions to retry on, possibly with their own retry settings, for something like #298 where you might not want to retry as much, but maybe at least once or twice in case there's some filesystem caching weirdness.