KinD cannot create cluster in rootless Podman #3234
Comments
|
have you followed the instructions in https://kind.sigs.k8s.io/docs/user/rootless/? |
|
Hello, thank you very much for quick response! I tried installing iptables-related modules manually: Next, I did with the same results. Well, the iptables-related complaints vanished, but the overall situation is the same. I'm attaching new logs: kind-logs2.zip |
|
containerd fails to create the pods maybe a problem with the storage? @AkihiroSuda does this ring a bell? |
|
Sorry, I forgot to mention that all my filesystems are XFS only. If that helps… |
This may work? |
|
With me I also can't upload the cluster with a normal user, I've tried everything and nothing. With sudo or ROOT it works normally. |
|
Hello friends! The runc#3805 is merged in master. It looks like it is not intended for the 1.1 branch. But anyhow. That could not stop me from trying. I built myself the master of runc at commit a6985522a6 and "patched" the official kindest:node like this:
Finally I created an cluster: and a single test shows the cluster is up-and-ready. When I take a look into events, there are only warnings (apart of normal ones) complaining about DNS: So far, so good. I hope this helps somebody. Thank you! |
|
Hello friends! Are there any plans or tactic how to get this resolved? Thank you. |
|
Sorry ... I don't work with podman regularly, the Kubernetes project requires docker to develop, so this is something we're looking for contributors to help maintain. It is very time consuming to debug issues with arbitrary linux environments. Thankfully, you've done that part, but it's stopped moving forward because this is only in run 1.2.x which is not released still. We take normal runc updates regularly. https://github.com/opencontainers/runc/pull/3805/commits => opencontainers/runc@da780e4 => This commit is only in the 1.2.x RCs, so it will be a while before we take it. We do not wish to make existing stable systems unstable. I would recommend using ext4 to run containers, especially if you're going to do container-in-container, there have been are a LOT of problems with detecting filesystem info, mounts, etc. that are not limited to code in this repo or runc and sticking to the most widely used tools (docker, ext4) is the most reliable path. You can see a number of other issues in the tracker where other filesystems caused issues for kubelet etc. |
What happened:
Hello friends,
As I have great success with KinD with Docker Desktop on Intel Mac, it has been my first choice to use it on my Linux box. Unfortunately I'm unable to create cluster.
I've been poking around and found an interesting issue, which has similar symptoms as I'm experiencing—#3061.
In short—details below—the
kind create cluster --config config.yaml -v 9999 --retainfails withI found out that kube apiserver is not running (hence the 6443 port is not listening).
There are two issues which caught my attention:
What you expected to happen:
The cluster spins-up and is ready to use.
How to reproduce it (as minimally and precisely as possible):
Anything else we need to know?:
config.yamlis used or not.The iptables excerpt
The permissions-related excerpt
kind-logs.zip
Environment:
kind version):docker infoorpodman info):/etc/os-release):kubectl version):The system uses Oracle Unbreakable Kernel instead of RedHat one:
Next, the system is switched from CGroupsV1 to CGroupsV2 with delegation/propagation. That works without issues in other containers.
Attached logs: kind-logs.zip
The text was updated successfully, but these errors were encountered: