Description of problem: in 3.10, the SCC admission plugin is skipped for components in "run level 0" namespaces (default, kube-system, etc), which means that components running in those namespaces do not have the allocated UID set on their pods. For components that use persistent storage, that means the uids of the pods do not match the file uid and read errors occur. Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. Run 3.9.x, with a registry and persistent storage. 2. Note the uid the registry pods run as. 3. Upgrade to 3.10.x (including updating the registry deployment to redeploy) 4. The registry pods encounter read errors on previously persisted data. 5. Note there is no uid forced onto the pods. Actual results: Expected results: Additional info:
fixed in https://github.com/openshift/origin/pull/19457
Created attachment 1437777 [details] api logs
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2018:1816