Description of problem: After upgrade of starter-us-east-2 yesterday evening it has been reported that we have started seeing the following issue. We are currently investigating the impact and will update the issue as we know more. 10:13:49 PM content-repository-1 Replication Controller Warning Failed create Error creating: pods "content-repository-1-" is forbidden: [minimum cpu usage per Pod is 17m, but request is 0., minimum cpu usage per Container is 17m, but request is 0.] 19 times in the last 14 minutes 10:13:48 PM jenkins-1 Replication Controller Warning Failed create Error creating: pods "jenkins-1-" is forbidden: [minimum cpu usage per Pod is 17m, but request is 0., minimum cpu usage per Container is 17m, but request is 0.] 19 times in the last Has there been a change to cpu requests with yesterdays upgrade with regards to minimum request limits? Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
Created attachment 1330954 [details] Logs across starter-us-east-2 tenants 260917
We had an account with a working jenkins that we reset the environment of and we are experiencing the same kind of issues on this one. This leads us to suspect that this might be the case for all accounts, or at least the ones with lack of limitrange?
The quickest solution for the workshop would seem to be to set DC limit requests.cpu to 17m instead of 0 (assuming the limitrange minimum cpu is consistent).
Putting a cpu request on a pod makes it subject to LimitRanges and ClusterResourceOverrides. If you want your pod best-effort, do not put a request at all. This can lead to poor performance on heavily loaded nodes, however, and increase risk for eviction. It is best to make your pod Burstable by setting a reasonable cpu request like 100m for a light/periodic task. This is a good reference to understand how/why/what to set your pod resource requests: https://docs.openshift.com/container-platform/latest/admin_guide/overcommit.html
Working as designed. Change in behavior caused by change in cluster configuration.