Eksctl: 0.1.2 release?

Created on 10 Sep 2018 · 6Comments · Source: weaveworks/eksctl

I'm currently getting an obscure error when running a create cluster command and I'm wondering if the changes since 0.1.1 will fix it. I searched for open issues around the unexpected status "ROLLBACK_COMPLETE" but didn't see anything yet.

It looks like there have been significant changes since the 0.1.1 release 10 days ago. Can we get a 0.1.2 release cut?

(Note: Foo name is edited in below to replace my real names.)

$ eksctl create cluster --name=Foo --node-type=m5.large --nodes=3 --nodes-min=3 --nodes-max=5 --ssh-public-key=foo-eks --region=us-east-1 --zones=us-east-1a,us-east-1b,us-east-1d
2018-09-10T14:30:41-04:00 [ℹ]  SSH public key file "foo-eks" does not exist; will assume existing EC2 key pair
2018-09-10T14:30:41-04:00 [ℹ]  found EC2 key pair "foo-eks"
2018-09-10T14:30:41-04:00 [ℹ]  creating EKS cluster "Foo" in "us-east-1" region
2018-09-10T14:30:41-04:00 [ℹ]  creating VPC stack "EKS-Foo-VPC"
2018-09-10T14:30:41-04:00 [ℹ]  creating ServiceRole stack "EKS-Foo-ServiceRole"
2018-09-10T14:31:22-04:00 [✔]  created ServiceRole stack "EKS-Foo-ServiceRole"
2018-09-10T14:32:04-04:00 [✔]  created VPC stack "EKS-Foo-VPC"
2018-09-10T14:32:04-04:00 [ℹ]  creating ControlPlane stack "EKS-Foo-ControlPlane"
2018-09-10T14:32:30-04:00 [ℹ]  an error has occurred and cluster hasn't beend created properly
2018-09-10T14:32:30-04:00 [ℹ]  to cleanup resources, run 'eksctl delete cluster --region=us-east-1 --name=Foo'
2018-09-10T14:32:30-04:00 [✖]  unexpected status "ROLLBACK_COMPLETE" while creating CloudFormation stack "EKS-Foo-ControlPlane"
2018-09-10T14:32:30-04:00 [ℹ]  creating DefaultNodeGroup stack "EKS-Foo-DefaultNodeGroup"

Running the suggested delete command also errors out giving a warning about the kubeconfig file and I'm not sure that it's always cleaning up after itself in CloudFormation / VPC.

$ eksctl delete cluster --region=us-east-1 --name=Foo
2018-09-10T14:41:06-04:00 [ℹ]  deleting EKS cluster "Foo"
2018-09-10T14:41:12-04:00 [!]  as you are not using the auto-generated kubeconfig file you will need to remove the details of cluster Foo manually
2018-09-10T14:41:12-04:00 [✔]  all EKS cluster "Foo" resource will be deleted (if in doubt, check CloudFormation console)

Note: I'm not sure about what's wrong with the kubeconfig file. I have an existing kubeconfig file from using Docker for Mac and GKE. If eksctl is adding its own entry or modifying the file or something, I haven't changed anything there related to eksctl myself.

When I run kubectl config get-contexts and kubectl config get-clusters, I don't see anything related to EKS or eksctl in the output.

Source

tedmiston

All 6 comments

I just found this in the CloudFormation logs under view failure event details.

15:07:45 UTC-0400 | CREATE_FAILED | AWS::EKS::Cluster | EKSCluster | Cannot create cluster '*' because us-east-1e, the targeted availability zone, does not currently have sufficient capacity to support the cluster. Retry and choose from these availability zones: us-east-1b, us-east-1c, us-east-1d (Service: AmazonEKS; Status Code: 400; Error Code: UnsupportedAvailabilityZoneException; Request ID: *)

I think I'm just experiencing the availability zones issue on the AWS side that's already documented (like https://github.com/weaveworks/eksctl/issues/75), not something with eksctl. I misread the original docs here and was hardcoding to the zones string in the docs.

tedmiston on 10 Sep 2018

Yes, this is availability issue that occurs in us-east-1, but had been seen in us-west-2, so I recommend using us-west-2.

The issue that we should fix is #190. Actually, what happen is that with the switching to CloudFormation in 0.1.1 (#126), we non-intentially hidden the error message away.