Tao Yang created YUNIKORN-410:
---------------------------------
Summary: Pod state change may cause incorrect update on
SchedulerNode#occupied
Key: YUNIKORN-410
URL: https://issues.apache.org/jira/browse/YUNIKORN-410
Project: Apache YuniKorn
Issue Type: Bug
Components: shim - kubernetes
Reporter: Tao Yang
Recently we found pod state change may cause incorrect update on
SchedulerNode#occupied. For example, if pod state changed from Running to
Pending then back to Running, after that SchedulerNode#occupied turned to
incorrect since it had added resource of the same pod twice.
The referenced code is in nodeResourceCoordinator#updatePod,
SchedulerNode#occupied will always be updated when pod state is changing to
Running, even if the pod has already been added to the cache of node. A simple
solution is to confirm that the new running pod has not been included in the
cache of node at first.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@yunikorn.apache.org
For additional commands, e-mail: dev-h...@yunikorn.apache.org
