GSA-TTS / datagov-brokerpak-eks

Broker AWS EKS instances using the OSBAPI (eg from cloud.gov)
Other
9 stars 7 forks source link

Add dependency for sc on cluster-functional #77

Closed nickumia-reisys closed 2 years ago

nickumia-reisys commented 2 years ago

In case kubenetes tries to create the storage class before the cluster is actually functional (i.e. node groups coming up or some other race condition), wait until we know the cluster is more ready

Related to,

Error: Error: Post "https://3ecc6d85ff42f6acfd57348528412984.gr7.us-west-2.eks.amazonaws.com/apis/storage.k8s.io/v1/storageclasses": dial tcp 100.20.251.29:443: i/o timeout  with kubernetes_storage_class.ebs-sc,  on persistent-storage.tf line 209, in resource "kubernetes_storage_class" "ebs-sc": 209: resource "kubernetes_storage_class" "ebs-sc" { exit status 1
mogul commented 2 years ago

@nickumia-reisys I've tested all of this from my local machine and it's working fine. However it seems like the $planid variable isn't populating under GitHub Actions, and I don't know why. Super weird, because it clearly works if I run make check-ids locally. Have any ideas?

mogul commented 2 years ago

OK, looks like it was only working locally because I had previous set serviceid and planid in my shell when I was testing... When I unset them and run make check-ids then I'm getting the same result as GitHub Actions. This means those env vars are not properly getting propagated between the lines where they're eval'd and the following lines.

The next step is to get make build up in a fresh shell where only the AWS creds are set, then fiddle with make check-ids to figure out something that works. I suspect this is going to mean turning the steps for each target into one long &&-compounded line.

Too tired to continue... Take a look and I will pick this up in the morning if you haven't!

mogul commented 2 years ago

Great stuff here.

nickumia-reisys commented 2 years ago

There's no reason that test should have failed to provision because it provisions locally for me just fine. It still has issues being destroyed, but that's not why the test failed either..