aws-node-termination-handler: Bad images on Public ECR?
Just started getting standard_init_linux.go:228: exec user process caused: exec format error
on the v1.13.3 image from the public ECR repo, which looks to have just been pushed an hour ago. This is on a standard x86_64 host (Amazon Linux 2).
Bad image example: public.ecr.aws/aws-ec2/aws-node-termination-handler:v1.13.3 sha256:63faf9b7bd275d505d150a0510813a9cb0cab9ad599a2b29d7cd70bd97209007
From the ECR gallery page (https://gallery.ecr.aws/aws-ec2/aws-node-termination-handler), a load of images were just pushed but luckily NOT the v1.13.4 image (sha256:259100e8c555e97170f0cb83578f2e5302d062d29e58ea3976ce54efce62f35b) which looks OK so far, and hence I’ve switched to that as a workaround.
About this issue
- Original URL
- State: closed
- Created 3 years ago
- Reactions: 18
- Comments: 24 (9 by maintainers)
The bad manifests in the ECR public repo have been patched. Apologies for the inconvenience this caused. Let us know if anyone is still experiencing issues.
We are facing the same issue.
For now, we retrieved the image digest from a working node and we set it on the deployment directly (instead of using the image tag).
Example for the
v1.12.2
version:For helm, we can trick the release by using
I also faced this issue. Upgraded to 1.14.0 and It is fine.
Any news? we are receiving the same error with v1.13.0
Thanks, I see new pods are starting ok now.
All manifests have been patched. Let us if you are still having problems. NOTE: you may need to change the image pull policy to
Always
since the default isIfNotPresent
which would cache the bad image pulled. Or you can roll the node.@kanor1306 v1.14.0 is available: public.ecr.aws/aws-ec2/aws-node-termination-handler:v1.14.0
So are we. v1.13.3