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
Clean up dockershim flags in the kubelet #106893
Comments
/cc @adisky |
/assign |
@cyclinder yes we need to keep
|
@adisky so we should marked |
lets handle it as a separate issue and leave it for now |
thanks @adisky ,I will work on this. |
@adisky Should |
@hakman It makes sense to remove |
removing the for 1.24 i suggest:
|
Sounds good to me @adisky @neolit123. I was planning to do it in a similar way. Will open a PR for it tomorrow. |
sorry @hakman I saw the reply just now created the PR #107094 |
@adisky I was thinking of removing the code behind it too, to make it a no-op. As is now, it would create more confusion. What do you think? |
#114017 is opened for this. |
Strange as it may seem, he is removing code that discards flags using fs.MarkDeprecated |
The pr author updated the pull request. |
Update some current status about this issue(half done):
|
The warning seems to only work with containerd, since the "config" field is not available in docker or cri-o ?
{
"status": {
"conditions": [
{
"type": "RuntimeReady",
"status": true,
"reason": "",
"message": ""
},
{
"type": "NetworkReady",
"status": false,
"reason": "NetworkPluginNotReady",
"message": "docker: network plugin is not ready: cni config uninitialized"
}
]
},
"config": {
"sandboxImage": "registry.k8s.io/pause:3.9"
}
} |
You can remove the warning in kubeadm now, since the kubelet will ignore the flag anyway (since v1.29.0-alpha.1). commit d5690f1 The actual flag will not be removed until 1.30, but it is not actually doing anything now (it now uses "pinned" instead) |
which k/k pr did that?
we can remove the flag EDIT: kubeadm can continue showing this warning: |
i can send the kubeadm pr for this. |
actually we are facing a bit of an issue after the recent change to support N-3 kubelet upgrades. please see: ideally, the flag should be removed after 1.29 goes out of support (if my math is correct). |
|
the flag cannot be removed in 1.30 if we follow the rules: |
/cc @mrunalp @SergeyKanzhelev |
/unassign @cyclinder |
no need to revert #118544. the flag must remain as a no-op for a GA deprecation period. not just one release. |
This line should be changed if I understand correctly. kubernetes/cmd/kubelet/app/server.go Lines 203 to 204 in 6897b12
kubernetes/pkg/kubelet/config/flags.go Line 54 in 6897b12
|
the deprecation timeline in the note can be adjusted if SIG node agrees. |
Now we can see the kubelet starting to garbage collect the sandbox image, in Docker...
(it gets a bit clean-happy, when you run kubernetes-in-container and it sees host Fortunately the cleaning failed, since there were still some kube-system pods using it. |
/assign |
From containerd side, containerd v1.7.3 and containerd 1.6.22 includes the support of pinned image:
However, there are still two bugfix pull requests in progress: |
Are those PR's really a blocker? one of them is for adding the pinned label for preloaded image which can be done via ctr as well while importing the image. Kind is planning to leverage it without those PR's, |
The two WIP PR are not blockers. To use that feature, we at least should use containerd v1.7.3+ and containerd 1.6.22+. |
/sig node
/kind cleanup
/cc @neolit123
As part of the removal of dockershim, we should also remove the deprecated CLI flags that still exist.
The
pod-infra-container-image
andcontainer-runtime
flags are now also deprecated and planned to be removed int v1.27.pod-infra-container-image
is being integrated through CRI implementations to avoid needing double specification.https://github.com/kubernetes/kubernetes/blob/master/pkg/kubelet/config/flags.go#L92-L112
The text was updated successfully, but these errors were encountered: