bci_repo_publish.py: Return 1 if the last published build is over a week old#2998
bci_repo_publish.py: Return 1 if the last published build is over a week old#2998Vogtinator wants to merge 1 commit into
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #2998 +/- ##
=========================================
Coverage ? 28.45%
=========================================
Files ? 85
Lines ? 14716
Branches ? 0
=========================================
Hits ? 4188
Misses ? 10528
Partials ? 0 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
Vogtinator
force-pushed
the
bcirepopublisher
branch
from
294f670 to
5ef36d9
Compare
…eek old This way it's easy to spot whether a manual look is required.
Vogtinator
force-pushed
the
bcirepopublisher
branch
from
5ef36d9 to
9f6b77f
Compare
|
I think we should simply stop publishing when there are failed builds? I'm not sure what hte cutoff of 7 days is solving? |
Currently the bot fails only if some API call actually fails, but not if a build failed or openQA test failed, as this is just expected behaviour. This means that if the repo is unresolvable for a month, the only sign is that there haven't been any openQA runs recently. Or if openQA fails for a month because nobody had a look, the jobs are still green. With this change it's visible in botmaster as a pipeline failure if there hasn't been a new build published for a week. |
| # Trigger publishing | ||
| if token is None: | ||
| self.logger.warning('Would publish now, but no token specified') | ||
| self.last_publish_mtime = packages[0]['built_mtime'] |
There was a problem hiding this comment.
why are we setting last_publish_mtime in this error case?
There was a problem hiding this comment.
It's not really an error case, more of a dry run case
There was a problem hiding this comment.
sure, but there are plenty of other returns before hand that are not setting last_publish_mtime. so this is a bit inconsistent.
| self.logger.info(f'{build_prj}/images not successfully built') | ||
| pkg = packages[0] | ||
| self.last_publish_mtime = \ | ||
| self.mtime_of_product(pkg['publish_prj'], pkg['name'], 'images', 'local') |
There was a problem hiding this comment.
and here? we're not actually publishing, we're aborting because it wasn't successfully built?
There was a problem hiding this comment.
Yes, which means we have to fetch the last published mtime from OBS to know how long ago that was
There was a problem hiding this comment.
as a nit pick, people usually avoid escaping newlines in python. you can write this in a more pythonic way:
self.last_publish_mtime = self.mtime_of_product(
pkg['publish_prj'], pkg['name'], 'images', 'local')There was a problem hiding this comment.
IMO harder to read and flake8 didn't complain 🤷
|
so if I understand this correctly. the only goal of the whole logic is to make the checkbox "red" rather than "green" in the botmaster listing/graph. how about we change the method to return True/False on whether it actually did a publish or not and then simply convert that to an exit code of 0 / 1? then it would be red if it had something to publish but didn't publish (because tests failed or no matching run found or publish failed), and otherwise green? does the UI support tripple state? then we could encode the "still waiting" states with another color.. |
Yes.
This wouldn't be able to catch if there was no new build for a week, e.g. due to unresolvable or blocked.
There's only orange for "pipeline is running" |
well, unresolvable or build failed would lead to an exit 1, as well as a missing openqa run. unfinished openqa run or unclean build status would not.
ok, too bad. Well, I don't have strong feelings either way. I think the 7 days to wait is arbitrary and feels a bit too long. maybe 3 days? |
3 participants
This way it's easy to spot whether a manual look is required.
I'm not entirely sure whether this is a good idea. Ideally someone looks directly when builds or tests fail and fixes this before a week expires, but as additional safe guard it's probably worth it.