You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
e2e(kafka): the suite leaves about 40 GiB of broker PVCs behind in tenant-test #4616
The kafka e2e suite leaves its broker PVCs behind in tenant-test, about 40 GiB per run, and later suites in the same sandbox pay for it.
Both node pools set deleteClaim: false (packages/apps/kafka/templates/migration-hook.yaml), which is intended: the README says to delete the volumes by hand after removing the app. hack/e2e-chainsaw/kafka/chainsaw-test.yaml deletes the app but never the volumes. In nightly run 36659663201 the PV snapshot taken after the kubernetes suites still shows them Bound: data-0-kafka-test-b-*-0/1 at 10Gi each and data-0-kafka-test-c-*-2/3/4 at 5Gi each, created at 03:03. 20 GiB of that sits on srv2, which is what kept srv2 below the replicated-lane pool floor and failed both kubernetes suites in that run (#4617).
Fix shape: the suite deletes the PVCs in its own cleanup, e.g. by the strimzi.io/cluster label for each test cluster, and waits for them to go.
Activity
fix(e2e): hold replicated-lane LINSTOR cleanup to a pre-suite baselin…