kind cluster on WSL2 does not restart when Docker restarts #3244
Comments
|
We're going to need logs from when it doesn't start to see why it exited or if docker even started it
|
|
But generally speaking short of a docker bug failing to restart it for some reason, I'd expect the entrypoint failed and that happens when something is rather broken in the environment and it is by design that it doesn't constantly keep restarting if that happens. |
|
Please also see the pinned issue about WSL2: #3180 |
|
|
|
@BenTheElder thank you! I collected some outputs (see above), hope that helps. |
|
Unfortunately these logs show that the container is running without exiting? We need the logs / exit status from when it fails to start running. I think some of the issues are fixed by either #3180 or #3241. For the latter it's not released yet and technically a breaking change but could be tested with https://kind.sigs.k8s.io/docs/user/quick-start/#installing-from-source |
Not sure, the problem is that container is not started on Docker Desktop restart no matter which So, technically, it could be logs from previous normal run, but they might not reflect the reason of |
So, if it never started at all following host restart, that's not a kind bug, that's a bug in docker desktop or a behavior change in docker specifically. But before saying docker desktop has a bug, I would guess that it could have started and failed, because the policy is to only start once. If it is started once and failed, we need the logs from that. Otherwise we need evidence that docker has broken this behavior and won't fix it before we can commit to changing our approach to something like always with the drawbacks that entails. |
|
Are we sure these logs are immediately after docker desktop was restarted without manually starting the kind cluster again? Because these logs appear normal, the container is in running state. |
BenTheElder
added
the
triage/needs-information
|
Here is what I have immediately after restart (other containers running as expected): |
OK, this is with the modified container settings though because kind does not set always. The logs don't make sense, maybe docker/for-linux#293 Have you checked the WSL version as mentioned in #3180 ? |
|
The original post says:
But as far as I know, latest WSL2 is known to be broken on their end, in the pinned issue linked here. |
Yes, default didn't work so I tried all of them and left "always".
It's |
|
1.2.5.0 is known to be broken per follow-up from another user (and updated title etc), can you try the steps there to switch to 1.2.4.0? 1.2.0.0 was also broken, and there was no change on our end, just a fix in WSL2. |
|
Interesting observation: sometimes |
|
@BenTheElder after several attempts I got this error message: and after a couple retries kind has been started w/o error message. Can this help? |
|
This is interesting, but can we confirm if it still happens on WSL2 1.2.4.0? As mentioned previously we have multiple reports of WSL2 shipping broken 1.2.X releases and 1.2.4.0 is considered working but 1.2.5.0 and 1.2.0.0 at least are not. There are instructions in the linked and pinned issue. |
What happened:
Hey! I have created a cluster on Ubuntu@WSL2, but found that it's not restarted when Docker restarts.
In order to start the cluster I have to run manually:
I have also tried changing
RestartPolicytounless-stoppedandalways- it didn't work. There are a couple of containers withunless-stoppedpolicy, and they all restart as expected, problem is only withkind-control-plane.When I run
docker start kind-control-planeeverything seems fine, so the problem is only with restart, I don't see any failure.What you expected to happen:
Kind cluster being started with Docker restart.
How to reproduce it (as minimally and precisely as possible):
If I'm correct:
Anything else we need to know?:
Environment:
kind version):docker infoorpodman info):/etc/os-release):kubectl version):No
The text was updated successfully, but these errors were encountered: