-
Notifications
You must be signed in to change notification settings - Fork 3k
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
Fix 'node-init' in GKE's 'cos' images. #19017
Conversation
|
/test Job 'Cilium-PR-K8s-1.21-kernel-5.4' failed: Click to show.Test Name
Failure Output
If it is a flake and a GitHub issue doesn't already exist to track it, comment |
3e2f9f9
to
a664df3
Compare
Commit 0bcb736535ff497260806efc36f11578bab2a3b3 does not contain "Signed-off-by". Please follow instructions provided in https://docs.cilium.io/en/stable/contributing/development/contributing_guide/#developer-s-certificate-of-origin |
Turns out that in GKE's 'cos' images the 'containerd' binary is still present even though '/etc/containerd/config.toml' is not. Hence, the kubelet wrapper would still be installed for these images according to the current check, even though it's not necessary. What's worse, starting the kubelet would fail because the 'sed' command targeting the aforementioned file would fail. This PR changes the check to rely on the presence of the '--container-runtime-endpoint' flag in the kubelet, which is probably a more reliable way of detecting '*_containerd' flavours and only applying the fix in these cases. Fixes cilium#19015. Signed-off-by: Bruno M. Custódio <[email protected]> Co-authored-by: Alexandre Perrin <[email protected]>
0bcb736
to
b5f3479
Compare
/test |
Turns out that in GKE's
cos
images thecontainerd
binary is still present even though/etc/containerd/config.toml
is not. Hence, the kubelet wrapper would still be installed for these images according to the current check, even though it's not necessary. What's worse, starting the kubelet would fail because thesed
command targeting the aforementioned file would fail.This PR changes the check to rely on the presence of the
--container-runtime-endpoint
flag in the kubelet, which is probably a more reliable way of detecting*_containerd
flavours and only applying the fix in these cases.Fixes #19015.