Skip to content
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

Unable to install eks anywhere with kindnetd cni due to image tags in public.ecr.aws have been changed. #7264

Open
eugenejen opened this issue Jan 5, 2024 · 12 comments
Labels
external An issue, bug or feature request filed from outside the AWS org
Milestone

Comments

@eugenejen
Copy link

What happened:
kindnetd image tag has been changed on public.ecr.aws from v0.18.0 to v0.18.0-eks-a-45

What you expected to happen:
This cause issue to install eks anywhere on docker when cillium cni has issue to start.

How to reproduce it (as minimally and precisely as possible):
on clean docker factory setting.
run

CLUSTER_NAME=mgmt eksctl anywhere generate clusterconfig $CLUSTER_NAME \ --provider docker > $CLUSTER_NAME.yaml

Then update mgmt.yaml from cilium to kindnetd
Then run
eksctl anywhere create cluster -f $CLUSTER_NAME.yaml

then use docker ps to inspect hang installation and you will see the image was unable to be pulled down for cni.

Anything else we need to know?:

Environment:

  • EKS Anywhere Release: 0.18.3
  • EKS Distro Release: 1.28
@vivek-koppuru vivek-koppuru added the external An issue, bug or feature request filed from outside the AWS org label Jan 7, 2024
@vivek-koppuru vivek-koppuru added this to the oncall milestone Jan 7, 2024
@robertlcx
Copy link

I'm also seeing the exact same issue. I'm trying to use Kindnetd because I cannot bootstrap the cluster with Cilium networking on my M1 Mac.

@jaxesn
Copy link
Member

jaxesn commented Jan 8, 2024

Ill take a look at this, I can repro it as well.

@robertlcx you should be able to create clusters with cilium on M1 Macs, I usually work on M1 mac as well.

@eugenejen
Copy link
Author

@jaxesn just curious. I am using docker for mac on intel. But I see there seems issue to run ebpf on docker for mac? The reason we use kindinetd cni is I can get kind running. but i am having issue to run cilium on docker itself and also in eks anywhere (cilium just crashed).

It there any setting we need to adjust in docker for mac desktop to enable ebfp?

@jaxesn
Copy link
Member

jaxesn commented Jan 9, 2024

Oh thats odd, I don't believe so. I will try to today on my M1 with cilium to make sure that still works as expected. Ill see if someone can try on an intel mac to confirm as well.

What version of docker do you have installed?

@eugenejen
Copy link
Author

@jaxesn I am using docker desktop for mac intel version

version 4.26.0 (130397)

Engine: 24.0.7

Compose: v2.23.3-desktop.2

Credential Helper: v0.7.0

Kubernetes: v1.28.2

@jaxesn
Copy link
Member

jaxesn commented Jan 12, 2024

I am seeing the same bpf issue on my mac as well. This is "newish", we've def seen this work in the past. Ill do a like poking around to see if there is a workaround.

@abhay-krishna also fixed the kindnetd manifest for our 0.18.x releases so you should be able to create docker clusters using kindnetd now.

@jaxesn
Copy link
Member

jaxesn commented Jan 12, 2024

I think this is the same issue: kubernetes/minikube#17780

Try downgrading to the 4.25.x release of docker for mac.

@abhay-krishna
Copy link
Member

@eugenejen @robertlcx were you able to get further in your cluster creation?

@eugenejen
Copy link
Author

@abhay-krishna i downgraded to 4.25.x. but now i am getting another issue on #6678 and unable to resolved it.

@abhay-krishna
Copy link
Member

@eugenejen did you try the workarounds suggested in that issue, particularly switching from VirtioFS to gRPC Fuse? You will find this option under the General tab in the Docker Desktop Settings menu.

@eugenejen
Copy link
Author

@abhay-krishna i have verified that I can swithc to use gPRC Fuse and the issue is resolved and cilium is running.

@robertlcx
Copy link

@eugenejen @abhay-krishna managed to fix this by downgrading to 4.25.x, switching back to cilium, and using gRPC Fuse instead.

Up until a couple of days, my fix was running an older version of eksctl and eks-anywhere, but now, some of the older images have been yanked from their Docker registry, so it wasn't working anymore.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
external An issue, bug or feature request filed from outside the AWS org
Projects
None yet
Development

No branches or pull requests

5 participants