wilfred-s commented on PR #521: URL: https://github.com/apache/yunikorn-core/pull/521#issuecomment-1486240210
As @craigcondit mentioned the normal allocation cycle properly handles any type of resource. If you set a limit on pods in the queue pods get limited after the change in apache/yunikorn-k8shim#558 The scale up issue is limited to the way we calculate the outstanding requests that fit in the headroom. I have been able to reproduce the issue in tests by simply adding pods that request a resource type that is not limited. That causes the headroom to go negative for certain resource types and nothing fits. I opened a [new bug](https://issues.apache.org/jira/browse/YUNIKORN-1658) to fix that side of the calculations and ignore unlimited resources as we should. Similar to the way we ignore them when we look at the fit on a node or in a queue. The k8shim side change for BestEffort pods should be tested on top of that change. This PR contains changes that could break current accounting of resources. The only points that can ignore undefined resources are in the `queue.internalHeadRoom()` and `application. getOutstandingRequests` as they look at quotas which are infinite if undefined -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
