This short report highlights some interesting commits to scylla-cluster-tests.git master from the last week. Commits in the d2a6d074…b3912362 range are covered.
There were 25 non-merge commits from 8 authors during that period. Some notable commits include:
Until now, each cleanup code made more API calls than necessary. Redundancy from the ‘clean-resources’ has been removed. This was causing issues with API call limits on the GCE side, we should no longer face this problem.
Previously, if the Jenkins pipeline failed before reaching the SCT test, it was not shown in Argus. This resulted in missing some failures and not raising issues. Now, we catch failed runs earlier.
The scylla-operator now supports monitoring. For some K8s tests, we install scylla-operator’s monitoring in K8S. This includes our SCT dashboard. Several commits were made to enable support for this feature, such as adding the possibility to ‘apply’ K8S files on the ‘server-side’, with the --server-side parameter for the kubectl apply operation. This allows us to apply large files that exceed the client-side limit and apply files ‘as is’ without needing to escape certain characters, which is problematic for ‘regex’ file lines.
Tests running in multi-az environments generate additional network traffic costs. We have added the possibility to simulate racks by tuning the GossipingPropertyFileSnitch snitch and provisioning instances in one availability zone to avoid them.
We look forward to the next issue of last week in scylla-cluster-tests.git master!