-
Notifications
You must be signed in to change notification settings - Fork 587
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
webhook CrashLoopBackOff with "Failed to start informers", "failed to wait for cache at index 4 to sync", when sources.knative.dev/v1beta1 SinkBindings exist #4876
Comments
Is it just with What about upgrades with the Operator (e.g. changing the |
Seems very much related to knative/operator#292 |
I don't think so, the original issue in knative/operator#292 was a lack of liveness probe initial delay (the time before liveness probe failure was shorter than the time it took to acquire the lease). In this case the informers will give up even if we lenghten the liveness probe delay, and it doesn't have to do anything with the lease (we can reproduce this one even if we delete the lease). |
@maschmid do you have a yaml for
? Not sure why I'd create |
@maschmid anything for a reproducer yaml or so ? |
I get
when I do step 3) after I deleted the |
With OpenShift serverless 1.13 , https://gist.github.com/maschmid/e0b04e9f4a6341ebf50d4076bb63a6c8 |
On eventing master, sinkbindings are stored as v1, so you cannot create v1beta1 without conversion (which will also mean that the issue is probably not present anymore on master, as the stored version and the version the webhook lists are the same, so the conversion is no longer attempted) |
master of knative/operator
On Thu 18. Feb 2021 at 13:00, Marek Schmidt ***@***.***> wrote:
On eventing master, sinkbindings are stored as v1, so you cannot create
v1beta1 without conversion (which will also mean that the issue is probably
not present anymore on master, as the stored version and the version the
webhook lists are the same, so the conversion is no longer attempted)
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#4876 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AABGPTT7NHL4RZJQTXC3VSTS7T6MFANCNFSM4XQLIZFQ>
.
--
Sent from Gmail Mobile
|
+1 sitting with a broken cluster as well, OCP 4.6.16, serverless operator 1.13.0. Is there anyway to get rid of the serverless component? |
To remove all v1beta1 sinkbindings on OpenShift,
If KnativeEventing is not installed, you may need to remove their finalizers as well.
|
OK, I've looked at this again. This does only happen if you use knative/operator (or distribution) 0.19.x WITH the However, using |
On Sinkbinding See:
|
The lisiting of the Sinkbindings Should that be kept at v1beta1 ? 🤔 |
Note in case it helps: this happened to me when a KafkaSource object had been created in a project without the CRD and the controller being deployed before (the user had skipped this step when configuring Serverless). |
I did some investigation here: knative-extensions/eventing-kafka#494 (comment) |
This issue is stale because it has been open for 90 days with no |
Describe the bug
When sources.knative.dev/v1beta1 SinkBindings already exist on the cluster during eventing-webhook startup, the webhook doesn't ever become ready in a CrashLoopBackOff, waiting for
eventing-webhook-certs
to be populatedIt seems that the informers go through the sinkbindings, but listing them calls the webhook itself (for conversion) before it's ready, which fails with
thus delaying the informers startup, which also delays the time it attempts to create the certs in
eventing-webhook-certs
, leading into a vicious cycle.Expected behavior
webhook should become ready even if old-versioned resources exists on the cluster.
To Reproduce
(with the Operator)
Knative release version
0.19.2
Additional context
The text was updated successfully, but these errors were encountered: