Last month in scylla-cluster-tests.git master (issue #39; 2024-03-10)

This short report brings to light some interesting commits to scylla-cluster-tests.git master from the last month. Commits in the da5bc7bd…516aa4d9 range are covered.

There were 60 non-merge commits from 11 Software Engineers in Test and 3 Software Engineers in that period. Some notable commits:

Readme manual got info about using AWS network configuration
refactored recently.

Nemesis can be skipped based on Scylla’s open issues, and we have a new command to scan the code for usages of SkipPerIssues and evaluate them if issues are closed and not tagged, so we can consider removing it from code, or tagging as needed.

There’s a new script for creating Jenkins pipelines based on directory structure. This is used to re-organize our pipelines, and it supports freestyle jobs and operator job creation. Also pipelines were split to two diffrent folders: oss and enterprise.

Since we run into multiple cases on parallel nemesis we introduced lock for target selection so the same node is not used multiple times in the same time.

We created a filter for tablets supported nemeses towards scylla version 6.0. To use it, use configurations/tablets_supported_nemeses.yaml in test configuration. Also introduced first tablets specific longevity and multi-dc multi-rack scenarios test.

EndOfQuotaNemesis was temporarily disabled due to an issue that won’t be handled meanwhile.

Instance types used in AWS tests were updated to the latest ones in most jobs.

We added support for streamlining OKTA usage this would check first if with default/used profile we can reach AWS, and if we don’t try to use gimmie-aws-creds tools to create a new profile.

New helper class was introduced to report versions of various tools used in SCT test runs in Argus. Currently, scylla’s python-driver version to both log and argus.

RestartThenRepairNodeMonkey nemesis was reenabled after a fix in Scylla.

New performance test with tablets enabled
measuring the latency during grow-shrink case for OSS and Enterprise versions. Disabled later due issues found by it.

Fixed weekly triggers use proper CI jobs and it’s parameters for operator tests and added a new weekly trigger for EKS ARM tests to cover ARM support in upcoming scylla-operator-v.1.12 release.

Raft feature is enabled by default for Scylla >= 5.5, in other cases it’s based on the consistent-cluster-management flag.

New upgrade test to tablets-enabled cluster. The test case starts with scylla 5.4 and during the upgrade process we activate tablets (and raft).

K8s tests use the newest scylla-manager version - 3.2.6. Also, we stopped specifying our own Scylla version which gets used as a backend for the scylla-manager and started to rely on the default value.

See you in the next issue of last week in scylla-cluster-tests.git master!