Last 2 weeks in scylla-cluster-tests.git master (issue #28; 2023-10-27)

This short report brings to light some interesting commits to scylla-cluster-tests.git master from the last 2 weeks. Commits in the 9d973065…63f27adc range are covered.

There were 40 non-merge commits from 10 authors in that period. Some notable commits:

Scylla log collected at the end of the tests with journalctl also has ms precision.. This log contains accurate time markers so are preferable for debugging.

We improved data validation for nemeses which delete data, so we have more confidence in data integrity after running them. For resources/time savings, we don’t verify all the data, just specified by configuration sample. Validation is done during cluster health check - in between nemeses. We can also now configure the test scenario to be more intensive on tombstones, by using delete_rows nemesis flag.

When using placement groups in performance tests, nodes added by nemesis are placed in the same placement group as other nodes.

In the ‘v1.11.0*’ scylla-operator was added the possibility to use PodIPs for node-to-node and node-to-client connectivity. We added support of direct usage of the PodIP addresses for DB pods.

We fixed and re-enabled the bare metal performance test.

New SCT ability to configure disks by custom scripts to cover custom use cases. Also, with use of the new scylla_d_overrides_files option in test configuration we can override any file in /etc/scylla.d folder. Scripts and custom configurations are stored in an external private repository, which can be pulled in the jenkins pipeline with checkoutQaInternal groovy helper.

Argus can show Gemini results similarly like shown in emails.

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