[ARCHIVED] CI is running parallel tests with undersized key buffers #2274
Labels
lifecycle/migrated-away
Indicates that this issue is archived and tracking continues in a copy.
priority/archived
Lowest priority possible - means that the issue is archived.
We are running 32 test in parallel in which a single ScyllaCluster needs several keys to be generated, yet we configure the buffers to:
https://gcsweb.scylla-operator.scylladb.com/gcs/scylla-operator-prow/pr-logs/pull/scylladb_scylla-operator/1955/pull-scylla-operator-master-e2e-openshift-aws-parallel/1868673877073203200/artifacts/deploy/0/operator/50_operator.deployment.yaml
This is barely enough to use the cache for one cluster and given how underprovisioned the operator is in the CI we can't afford to generate all keys on demand, so we should use the cache as much as possible to utilize the free cycles we have.
This is an example timing from a test that failed because cert were generated on demand for multiple test and too slow
This likely cause many e2e tests to flake.
The text was updated successfully, but these errors were encountered: