-
Notifications
You must be signed in to change notification settings - Fork 40k
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
Remove the flannel experimental overlay #33862
Remove the flannel experimental overlay #33862
Conversation
Jenkins GCE e2e failed for commit c99f586. Full PR test history. The magic incantation to run this job again is |
Jenkins GCI GCE e2e failed for commit c99f586. Full PR test history. The magic incantation to run this job again is |
Jenkins GCE Node e2e failed for commit c99f586. Full PR test history. The magic incantation to run this job again is |
I can also remove the proposal if you like |
@luxas you need to run |
Yes, I know :) @yujuhong PTAL at the first commit |
Fair enough. It does take quite a while to run. |
@yujuhong What do you think about the PR itself? |
Salt needs some cleanup too:
|
@luxas Thanks for the PR! Just the quota exceeded.
|
Jenkins verification failed for commit b251dd5. Full PR test history. The magic incantation to run this job again is |
I'm fine with removing this flag, and in general happy to see dead code go. Someone from the cluster team may want to chime in. |
Should I remove the old proposal as well: https://github.com/kubernetes/kubernetes/blob/master/docs/proposals/flannel-integration.md? |
I'm testing now to remove the old flannel code in kube-up as well. cc @roberthbailey @mikedanese @zmerlynn from the cluster team, PTAL |
containerManager: kubeDeps.ContainerManager, | ||
nodeIP: net.ParseIP(kubeCfg.NodeIP), | ||
clock: clock.RealClock{}, | ||
os: kubeDeps.OSInterface, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
why changing this?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It's just gofmt. I removed the two flannel overlay-specific fields
Jenkins Kubemark GCE e2e failed for commit a6ab75d. Full PR test history. The magic incantation to run this job again is |
Jenkins GCI GKE smoke e2e failed for commit a6ab75d. Full PR test history. The magic incantation to run this job again is |
Jenkins GKE smoke e2e failed for commit a6ab75d. Full PR test history. The magic incantation to run this job again is |
I don't know the policy on proposals, but I think removing a no-longer-relevant proposal is good. |
@@ -121,7 +121,7 @@ func (kl *Kubelet) providerRequiresNetworkingConfiguration() bool { | |||
// is used or whether we are using overlay networking. We should return | |||
// true for cloud providers if they implement Routes() interface and | |||
// we are not using overlay networking. | |||
if kl.cloud == nil || kl.cloud.ProviderName() != "gce" || kl.flannelExperimentalOverlay { | |||
if kl.cloud == nil || kl.cloud.ProviderName() != "gce" { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we still want a flag that does this probably. flannelExperimentalOverlay is the wrong name.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
One can use the CNI flannel plugin. But that needs extra setup for flannel daemon and etc.
I think that is much cleaner than having a flag for flannel.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
At least, this isn't a regression.
Nobody (at least not many) has used the experimental flannel overlay, and if we want to add a new flag for this specific case, we can do it later
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@freehan I think we have the same problem for not just flannel. If I'm using the gce cloud provider with weave and I don't want to configure routes, kubelet will report not ready. There is no way of opting out of gce routes with the gce cloud provider. Is that an issue?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@mikedanese what does kubelet complain in the weave case? I do not recall kubelet checks podCIDR unless it comes from network plugin. I think the assigned route can be ignored by configuration.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think you should open a separate issue for that discussion.
Please do that and keep focusing on the PR itself
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I brought this up in #33573 (comment). It's a problem for OpenShift when run on GCE too, as we use an overlay and don't need Routes created for pod<->pod communication, but the NetworkUnavailable condition is set anyway.
But I think this should be solved differently and more cleanly than repurposing the flannelExperimentalOverlay flag.
No. You can just leave it as is or make a not that it's deprecated. |
LGTM |
LGTM. Please rebase. |
Feel free to rebase and self apply |
a6ab75d
to
4e52f84
Compare
Jenkins unit/integration failed for commit 4e52f84. Full PR test history. The magic incantation to run this job again is |
@k8s-bot unit test this |
@k8s-bot test this [submit-queue is verifying that this PR is safe to merge] |
Automatic merge from submit-queue |
What this PR does / why we need it:
It removes the deprecated flannel overlay integration in kubelet.
It's completely unnecessary now with CNI which can handle everything smoothly when flannel is running in a daemonset.
Which issue this PR fixes (optional, in
fixes #<issue number>(, #<issue_number>, ...)
format, will close that issue when PR gets merged): fixes #17795, #30589cc @kubernetes/sig-network @thockin @freehan @bprashanth @yujuhong @dchen1107
This change is