TingYao Huang created YUNIKORN-3471:
---------------------------------------
Summary: Flaky priority_scheduling e2e:
Verify_Static_Queue_App_Scheduling_Order times out waiting for the high
priority pod
Key: YUNIKORN-3471
URL: https://issues.apache.org/jira/browse/YUNIKORN-3471
Project: Apache YuniKorn
Issue Type: Bug
Components: test - e2e
Reporter: TingYao Huang
Intermittent failure, e.g. e2e-tests (v1.35.8) on k8shim PR #1100
(https://github.com/apache/yunikorn-k8shim/actions/runs/35518564590/job/106110759015).
Passes on the other K8s versions and on rerun.
[FAIL] PriorityScheduling [It] Verify_Static_Queue_App_Scheduling_Order
priority_scheduling_test.go:452 — Unexpected error: context deadline exceeded
Setup: root.fence caps at one pod, child1 = low + high, child2 = sleep + normal.
The test deletes the blocking sleep pod (child2) and expects high (child1) to
take the
freed slot. In the failing run normal (child2) grabbed it instead, so high
stayed
pending until the 30s timeout.
Looks like a race when a freed slot is filled across sibling queues, not a
consistent
mis-sort. Is strict cross-queue priority expected here, or should the test be
hardened?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
