Learn more about these different git repos.
Other Git URLs
if the compose fails really early for things such as config validation we need to emit a message that the compose failed. Likely we should make setting up the compose layout on disk which locks in the compose id and emiting a message that the compose has started before doing config validation, if validation fails or something else really early on fails we need to emit the failure of the compose so that people can be notified when an automated compose has failed.
There is too many automated failures that people are not paying attention too. as we automate more its going to become more and more necessary.
Adding my 2 cents here - request to do this as soon as reasonable so that our automation is able to run reliably.
This is a little tricky: we don't know what the ID is going to be without the configuration. There will always be room for failure with no message sent.
PR #441 sends a message when CLI options are wrong or config is incorrect. It uses a new topic org.fedoraproject.prod.pungi.compose.fail.to.start (I'm open to suggestion for improvement). The downside is that it can not know the ID of the compose, so instead it gives the config file, target dir for the compose (/mnt/koji/compose/rawhide etc.) and the full command line that was called.
org.fedoraproject.prod.pungi.compose.fail.to.start
/mnt/koji/compose/rawhide
Example payload:
{ "config": "/home/lsedlar/repos/pungi/tests/data/dummy-pungi.conf", "target_dir": "/home/lsedlar/repos/pungi/tests/_composes", "errors": [ "Unrecognized config option: rlease_type. Did you mean release_type?" ], "command": [ "../bin/pungi-koji", "--notification-script=/home/lsedlar/repos/pungi/tests/../logger.sh", "--config=data/dummy-pungi.conf", "--target-dir=_composes", "--test" ], "detail": "Config validation failed" }
@lsedlar changed the status to Closed
Closed
Log in to comment on this ticket.