Closed Sponge-Bas closed 3 months ago
we'll do some digging. But can you confirm that when the hook retries the exact same exception occurs?
Solution will probably be to include ConnectionRefusedError in the catch group at https://github.com/canonical/cos-proxy-operator/blob/main/src/charm.py#L382
this might occur because we're trying to push stuff to loki before loki is ready. Question is: should we check loki is ready and if not, avoid trying? If we set Blocked and exit, when are we going to retry pushing the vector config?
The urlopen check is no longer in the code so shouldn't be an issue anymore.
Bug Description
In SQA testrun 651a309e-a3a6-44ab-b8a7-7905303fbc0a, cos-proxy fails to install in hook "downstream-logging-relation-changed".
To Reproduce
To reproduce, deploy cos and then the charmed kubernetes bundle, which includes cos. This issue is not necessarily reproducible, we have seen this bundle deploy without issues before.
Environment
The environment is a juju maas controller hosting a charmed kubernetes deployment. This deployment is connected to cos, which is hosted on a microk8s, hosted on the same juju maas controller.
Relevant log output
My guess is that there was a hiccup in the networking. If that is the case, though I would expect that to resolve when the hook is retried.
Additional context
More logs and configs can be found here: https://oil-jenkins.canonical.com/artifacts/651a309e-a3a6-44ab-b8a7-7905303fbc0a/index.html