[Existing VPC] : Tag subnets as 'owned'

[wiquan]

What steps did you take and what happened: With a pristine new stack in an existing VPC, much of the stack materializes, but some tags are missing on subnets such that any attempt to deploy something via Helm (eg. NextSteps:Wordpress ).

With an existing VPC and existing Subnets, you will hit:

• https://github.com/kubernetes/kubernetes/issues/29298#issuecomment-356826381

because there are tags that are missing. Thus, it would be good to either add the tags to the existing subnets or WARN users that they will be getting a situation like this:

# External IP is wedged as pending because tags are missing on subnets
$ kubectl get svc --namespace varmywordpress wordpress-wordpress
NAME                  TYPE           CLUSTER-IP      EXTERNAL-IP   PORT(S)                      AGE
wordpress-wordpress   LoadBalancer   10.104.13.150   <pending>     80:30789/TCP,443:32346/TCP   8m

What did you expect to happen: Expected the External IP to be replaced with an AWS resource. This happens if the template adds AWS Tags to the provided subnets for the stack being created.

Key= kubernetes.io/cluster/<<STACKNAME>>
Value= owned

$ kubectl get svc --namespace varmywordpress wordpress-wordpress
NAME                  TYPE           CLUSTER-IP     EXTERNAL-IP        PORT(S)                      AGE
wordpress-wordpress   LoadBalancer   10.100.207.5   abc12141d1b53...   80:31683/TCP,443:32357/TCP   24s

Anything else you would like to add: [Miscellaneous information that will assist in solving the issue.]

Environment:

• Kubernetes version: (use kubectl version):
• OS (e.g. from /etc/os-release):

[wiquan]

Subnets can also be tagged as 'shared' (instead of owned). That works better for me, since my single VPC has multiple stacks deploying into the same subnets.

[timothysc]

So we might totally revamp this once we long term switch to cluster-api implementation, under the hood.

[timothysc]

/assign @chuckha @vincepri