Due to BZ#1794714 We have added a MachineConfig to disable tx checksum offload on https://github.com/openshift/machine-config-operator/pull/1606. This needs to be removed when the Bug is resolved.
Why are we only doing this for workers? In compact clusters (no workers) masters would also need this, no?
(In reply to Scott Dodson from comment #1) > Why are we only doing this for workers? In compact clusters (no workers) > masters would also need this, no? This fix is to make NodePort work. NodePort is used to exposes services which should run on worker nodes. I know that you can use NodePort with masters as well but this is not the "usual/recommended use case" and disabling checksum by default is a workaround and not a solution. So we decided to apply this only to worker nodes
This looks like it merged. Should this possibly be MODIFIED for QE to verify? Or is this bug still being worked on?
Yes, I think so. The PR was never properly linked to the bug so the bug never transitioned to MODIFIED as it should have.
This bug is for tracking the removal of the workaround. The kernel bug should be resolved on kernel-4.18.0-202.el8 from [1]. Once we will have RCHOS with it the tests should pass and we can remove that workaround. It looks like it will be in on RHEL 8.3, so I will retarget to 4.6 and this should be on NEW since we are not removing it at the moment. [1] https://bugzilla.redhat.com/show_bug.cgi?id=1794714
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 (OpenShift Container Platform 4.6 GA Images), 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-2020:4196