Peter Toth created SPARK-59734:
----------------------------------
Summary: Remove the unreachable resource retain handling from
`ApplicationStatus.terminateOrRestart`
Key: SPARK-59734
URL: https://issues.apache.org/jira/browse/SPARK-59734
Project: Spark
Issue Type: Sub-task
Components: Kubernetes
Affects Versions: kubernetes-operator-1.1.0
Reporter: Peter Toth
After SPARK-59732, `AppCleanUpStep` is the only production caller of
`ApplicationStatus.terminateOrRestart` and it always passes
`ResourceRetainPolicy.Never`. That makes the retain handling inside the method
unreachable:
* the guard in the "no restart configured" branch,
* the guard in the "restart limit exceeded" branch,
* the private `terminateAppWithoutReleaseResource` helper they both call.
Every `TerminatedWithoutReleaseResources` state a running operator writes comes
from `AppCleanUpStep` itself, which keeps the retain decision in one place.
Deleting both guards and the helper leaves the whole `spark-operator` test
suite green; only the two `ApplicationStatusTest` cases that call the method
directly with `ResourceRetainPolicy.Always` fail, so the branch's only
remaining coverage asserts a state no production path can reach.
Proposed:
* delete the two guards and `terminateAppWithoutReleaseResource`,
* drop the now unused `resourceRetainPolicy` parameter of `terminateOrRestart`,
which is the part that changes a public signature in `spark-operator-api`,
* adjust the two `ApplicationStatusTest` assertions.
Note that `terminateAppWithoutReleaseResource` also concatenates its message
without a separating space, the same defect SPARK-59732 fixed in the live
branch. Deleting it resolves that too, so it should not be patched separately.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
