Description of problem: After a cluster upgrade from OCP 4.7.33 to OCP 4.8.26 a big increase in CPU load on the network switches has started to affect all connected entities. The reason seams to be related to the fact that the switches is discovering Service IP addresses on the node NIC. Version-Release number of selected component (if applicable): OCP 3.8.26 How reproducible: In the customer environment on several clusters. Steps to Reproduce: 1. Upgrade cluster from 4.7.33 to 4.8.26 2. Look at assigned IPs on the network 3. Actual results: The switch is detecting many service IPs Expected results: The switch should not detect service IPs Additional info:
Marking BZ with a depends on https://bugzilla.redhat.com/show_bug.cgi?id=2054394 since the fix really needs to go into OVN before we can consume it. This bug will continue to be open to track the OVN bump needed to consume this feature and to disable the GARPs for LB IPs on OVN-K side for egressIPs once the OVN fix lands. Note that this will take a few weeks.
Upstream fix: https://github.com/ovn-org/ovn-kubernetes/pull/2835
moving it back to post to pick up the second commit: https://github.com/ovn-org/ovn-kubernetes/pull/2835
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 (Important: OpenShift Container Platform 4.11.0 bug fix and security update), 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/RHSA-2022:5069