craigcondit commented on PR #521: URL: https://github.com/apache/yunikorn-core/pull/521#issuecomment-1481856558
There are two phases to scheduling. In the first phase, we only look at whether or not an ask fits into a queue -- in this case, the existing logic should be correct, as if a queue doesn't specify a limit, it is in effect unlimited. This goes for any resource type. The second phase deals with placement on nodes. Once pods:1 is added to each request, this should get handled properly by the node placement logic (as resources will not fit in that case). YuniKorn will only mark a pod as unschedulable (and therefore eligible for autoscaling) if it would fit in a queue, but NOT on any available node. TL;DR: I don't think this is the correct fix, as the existing logic should already do the right thing. @wilfred-s do I have this right? -- 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]
