Last month in scylla-cluster-tests.git master (issue #16; 2023-07-21)

This short(ish) report brings to light some interesting commits to scylla-cluster-tests.git master from the last ~month. Commits in the a1036e1b…9dfb2392 range are covered.

There were 116 non-merge commits from 15 authors in that period. Some notable commits:

As a first step in testing Audit functionality on larger scale we added short longevity with audit enabled by default.

SCT was adapted to uuid_sstable_identifiers_enabled where sstables are identified with timeuuid instead of an integer.

Check out docs: How can I connect to test machines in AWS ?

Since releng is moving to create the images on another account, we have introduced the changes to list the images regardless general or releng accounts but were missing some fixes and improvements like: add account images tags, add image’s owner ID, and checking owner ID.

New GH action was added to close stale issues after 2 years of inactivity and PR’s after 1 year, so we can enjoy fast and consistent pace of closing issues and PR’s.

This change adds a Global Secondary Index to a 3 hour large partition test and a materialized view to the 200k pks 4days test.

This change adds a random chance to use a “USING TIMESTAMP” clause instead of normal 50% of the partition in delete_by_range_using_timestamp disruption.

K8S 1.27 is now supported in SCT tests

Because the old ‘static’ approach for PV provisioning on GKE doesn’t work on latest K8S version, use dynamic local volume provisioner on GKE. Later static local volume provisioner was fixed using solution recommended by Google.

Introduced a new perf throughput non-shard aware i4i tests

K8s tests were swithed to use and test manager 3.1

New event process is presented and new context manager. Context manager allow to start/stop count events and collect some stats event process allow to filter collected events and save events to files in specified directory.

Nodetool cleanup nemesis was refactored to run in parallel without specifying keyspace.

In order to catch more issues with installations and upgrades that are related to instance type sizes (e.g. perftune issues related to amount of cores) we spreaded 5 common instances types in gce with higher amount of disks accordingly

Now we are able to set longevity test duration from job settings.

Hinted handoff feature was disabled for all tests as probably is rarely used on the field, leaving one sanity test to make sure it still works.

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