In order to test OCL we run e2e automated test cases in a cluster that has OCL enabled in master and worker pools.

We have seen that rarely a new machineconfig is rendered but no MOSB resource is created.




    

4.18
    

Rare
    

We don't have any steps to reproduce it. It happens eventually when we run a regression in a cluster with OCL enabled in master and worker pools.


    

We see that in some scenarios a new MC is created, then a new rendered MC is created too, but now MOSB is created and the pool is stuck forever.

    

Whenever a new rendered MC is created, a new MOSB sould be created too to build the new image.

    

In the comments section we will add all the must-gather files that are related to this issue.


In some scenarios we can see this error reported by the os-builder pod:


2024-12-03T16:44:14.874310241Z I1203 16:44:14.874268       1 request.go:632] Waited for 596.269343ms due to client-side throttling, not priority and fairness, request: GET:https://172.30.0.1:443/api/v1/namespaces/openshift-machine-config-operator/secrets?labelSelector=machineconfiguration.openshift.io%2Fephemeral-build-object%2Cmachineconfiguration.openshift.io%2Fmachine-os-build%3Dmosc-worker-5fc70e666518756a629ac4823fc35690%2Cmachineconfiguration.openshift.io%2Fon-cluster-layering%2Cmachineconfiguration.openshift.io%2Frendered-machine-config%3Drendered-worker-7c0a57dfe9cd7674b26bc5c030732b35%2Cmachineconfiguration.openshift.io%2Ftarget-machine-config-pool%3Dworker


Nevertheless, we only see this error in some of them, not in all of them.