This short report brings to light some interesting commits to scylla-cluster-tests.git master from the last week. Commits in the c18f651c…88ce7170 range are covered.
There were 26 non-merge commits from 7 authors in that period. Some notable commits:
Since manager 3.2.0 is being released, we bumped manager version to 3.2 in all tests.
Added a nemesis that disables the target node’s binary protocol and gossip, initiates a major compaction, and enables the node’s gossip and binary back. This scenario was found to be used on production and required more testing.
Using use_preinstalled_scylla: true as default in gce.
New microbenchmark test on x86_64 (c6i.large).
Recently the structure of defaults/manager_persistent_snapshots.yaml has changed drastically and mgmt_cli_test was adapted to it.
During upgrade tests when one node is upgraded, sometimes we observe read from truncated table to fail with timeout (60sec). We prolonged some queries timeouts and started to measure&store actual time in ES for future analysis.
Perf Emails subjects now contain the backend and node type for tests that support different backends. Find Prometheus stats to the bottom of perf emails body.
Operator v1.10 runs cleanups on each of the old nodes after adding a new one (like it is recommended by the Scylla documentation). We added a new functional test to cover that functionality.
See you in the next issue of last week in scylla-cluster-tests.git master!