Aditya Maheshwari created YUNIKORN-3316:
-------------------------------------------
Summary: Retry pod scheduling in case of failure during AssumePod
Key: YUNIKORN-3316
URL: https://issues.apache.org/jira/browse/YUNIKORN-3316
Project: Apache YuniKorn
Issue Type: Sub-task
Reporter: Aditya Maheshwari
Assignee: Aditya Maheshwari
Once the node for a pod is selected by core, it call rmcallback to update
allocation. During UpdateAllocation, shim tries to call AssumePod which updates
in-memory cache of k8s for pod volume bindings (not the actual binding). We
have retries in place to retry any failure during this step. But if all retries
get exhausted we fail the task and ask core to release the allocation.
As per current behavior pod should be in pending state in this case (to be
verified)
Instead of failing the task, we should retry again the task after all the
cleanup.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
