This short report brings to light some interesting commits to scylla-cluster-tests.git master from the last week. Commits in the 9c59cc58…da5f0d43 range are covered.
There were 29 non-merge commits from 6 authors in that period. Some notable commits:
SCT logs are now printing hostnames for each RemoteCommandRunner output line, which helps differentiate and filter log lines for specific nodes.
We introduced Amazon Linux 2023 tests as Amazon Linux 2 will be deprecated next year, so we are beginning support for the next version.
Another speed improvement for database node setup was achieved by moving syslogng-exporter installation to cloud-init.
Enabled testing of Kafka connectors with a Kafka cluster setup on sct-runner instance using Docker Compose. With a Kafka consumer thread, the test reads data written by the connector and validates the received information, starting with row count. Initial pipelines for the basic case were added, using Docker and AWS backends.
A small improvement for the apt packages installer to use a built-in feature to wait for locks.
When topology operations are aborted by SCT request, the severity of expected raft_topology error messages are decreased to a warning.
The large partition longevity test stress commands were refactored to simplify and behave correctly when triggered with a custom stress duration.
A new longevity test with RF=1 was added to cover issues observed in the field.
Rolling restart now waits for all nodes to be in UN state before proceeding to the next node.
Continued adaptation of Mergify to automate the backporting process within SCT.
Report screenshots are now created with Grafana Image Renderer, greatly improving the speed (from 9-10 minutes to 2.5 minutes). Lots of code could be dropped and selenium removed from the Hydra image.
See you in the next issue of last week in scylla-cluster-tests.git master!