Automated backport of #2399: Fix race when out-of-order RemoteEndpoint events are seen #2430
Conversation
When there is a gateway migration in a remote cluster or if there is any stale endpoint on the Broker associated with the remoteCluster, the events might come in out of order which can create issues for datapath connectivity. The existing code was already having checks to ignore stale endpoint notifications in the kubeproxy and OVN handler. However, similar check was missing in MTU handler and an issue was seen in one of the environments. This PR moves the checks to the event handler framework and removes them from the registered handlers so that the fix will apply to all the handlers. Signed-off-by: Sridhar Gaddam <sgaddam@redhat.com>
|
🤖 Created branch: z_pr2430/sridhargaddam/automated-backport-of-#2399-origin-release-0.13 |
|
@sridhargaddam any idea why E2E was failing here? |
Which one? |
mm I am pretty sure https://github.com/submariner-io/submariner/actions/runs/4731824505/jobs/8397222931?pr=2430 was failing, but is shown as "skipped" now? |
Okay, triggered a re-run. Lets see 👍 |
|
@dfarrell07 @tpantelis the E2e job is getting skipped. Any idea why? |
E2E Full - b/c it needs 2 approvers or ready-to-test 😄 |
|
🤖 Closed branches: [z_pr2430/sridhargaddam/automated-backport-of-#2399-origin-release-0.13] |
Backport of #2399 on release-0.13.
#2399: Fix race when out-of-order RemoteEndpoint events are seen
For details on the backport process, see the backport requests page.