-
Notifications
You must be signed in to change notification settings - Fork 2.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
CI: Fix Suite-runtime.RuntimeDatapathPrivilegedUnitTests #23297
CI: Fix Suite-runtime.RuntimeDatapathPrivilegedUnitTests #23297
Conversation
/test |
794a37f
to
db4b4a7
Compare
/test |
Can't get the tests better than this. The smoke tests error on the hubble relay container: |
Do we have an issue for that? I also see that the k8s-1.16-kernel-4.9 didn't run for some reason. |
We do now #23533
Oh, missed that. Perhaps we can re-trigger it, I will try |
/test-1.16-4.9 Job 'Cilium-PR-K8s-1.16-kernel-4.9' has 3 failures but they might be new flakes since it also hit 1 known flakes: #22056 (91.04) |
db4b4a7
to
185d663
Compare
/test |
185d663
to
8fdec44
Compare
/test |
8fdec44
to
93d60b5
Compare
/test |
We need to replace the t.Fatal calls with t.Error when within the await callbacks. These callbacks are called by the fake client but from the goroutine of the LBIPAM controller. This means that we panic in the controller causing us to not call `.Done` on resource events. Not calling `.Done` on resource events triggers the finalizer based checks to enforce `.Done` is always called. This masks the original error. By using t.Error we do mark the test as failed but only ever panic in the test suite itself. Signed-off-by: Dylan Reimerink <dylan.reimerink@isovalent.com>
This unit test was incorrect. It assumed that services with requested IPs whom didn't match the pool selector would be ignored. But this behavior was changed at some point during development so we instead set a status condition on the service. However the test wasn't updated. The test still returned PASS because it is time based and terminated after 100ms, but 7999/8000 times LB-IPAM takes longer than 100ms to send the update. This flake occurs only when we send the update in <100ms after start. The test has been updated to reflect the correct behavior, which surfaced a bug that was fixed as well. Fixes: cilium#23272 Signed-off-by: Dylan Reimerink <dylan.reimerink@isovalent.com>
93d60b5
to
d54f321
Compare
/test |
I finally got CI to agree with me on this one. We are still hitting #24217 but that is it. All reviews are in. CI green except for the flak on |
This fixes both the masking issue and the underlying issue, please see commit descriptions for details.
Fixes: #23272