Jack Francis
Jack Francis
@sbueringer so you're suggesting that the exit code result is < 100% reliable in terms of terminal outcome, and we need to fall back to introspecting junit whenever we get...
[This](https://github.com/kubernetes-sigs/cluster-api-provider-azure/issues/2265#issuecomment-1414533434) suggests that we (capi e2e) are getting the exit code from the conformance test.
In @CecileRobertMichon's linked repro prow *did* get a zero exit code: ``` + EXIT_VALUE=0 ``` @sbueringer This is why I wonder if your comment suggests the edge case we're after...
@sbueringer agreed In which case, I think we may be misconfiguring kubetest? The configuration surface area is so vast, perhaps the errors observed in these scenarios are configured to be...
Interesting, have you observed a repro since that change? CAPZ update to v1.3.3 in-progress: https://github.com/kubernetes-sigs/cluster-api-provider-azure/pull/3136
@CecileRobertMichon I'm not aware of any active investigation into this. I audited this page's history of successful results and didn't encounter any that exhibited the documented symptoms here. Perhaps we...
/test pull-cluster-autoscaler-e2e-azure
/lgtm /approve