Last week in scylla-cluster-tests.git master (issue #7; 2023-04-08)

This short report brings to light some interesting commits to scylla-cluster-tests.git master from the last week. Commits in the a3173984…4048674e range are covered.

There were 27 non-merge commits from 8 authors in that period. Some notable commits:

In order to verify correctness of gc-modes added new test for switching tombstone-gc modes in scaling and high stress.

Added new basic longevity for MV running in synchronous mode. It creates different materialized views. Part of MVs are running in sychronous mode, others in asynchronous mode. Updates/inserts/deletes into MVs perform in parallel with different nemeses, disruptive and non-disruptive

In case of newly added Scylla node failing on boot fail fast whole test, in that case test anyway was doomed to fail. Speeding up issues investigation.

New test options for Scylla on K8s allowing us to set cpu and memory limits for Scylla pods
(k8s_scylla_cpu_limit and k8s_scylla_memory_limit). When defined, will be applied to each Scylla pod. Also, these options are suitable for multitenant configuration.

Added 3 jenkins pipelines for longevity azure: longevity-1tb-5days-azure , longevity-200gb-48h-azure and longevity-large-partition-4days-azure .

We did some performance tests for Scylla serverless with 1 and 2 vcpus and these are the configs.

When testing, various operations take more time when node performance is lower or data load is higher. We aim to dynamically calculate timeout based on node performance and load. To enable it we started to measure how much time operations take along with node load and basic performance metrics. When we gather enough data, we will be able to calculate sensible timeouts dynamically.

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