do-not-return-deleted-nat-gateways

[calvix]

towards https://github.com/giantswarm/giantswarm/issues/14731

Checklist

• Update changelog in CHANGELOG.md.

[calvix]

tested by creating a cluster, deleting it, and recreating again, everything worked and the deleted NAT gateways wer ignored

[xh3b4sd]

Can you elaborate on how you create, delete, create the cluster? I would not know how to even do that. 😅

[xh3b4sd]

Also, AFAIK it is more dangerous to create clusters with the same ID twice due to the KMS encryption keys. That could have unintended side effects and we never really did or recommend that.

[calvix]

I simply created cluster via gsctl create cluster, then waited until cluster was up and ready.

Then dumped all the CRS into YAML file (AWSCOntrolPlane, G8sControlPlane,AWSCluster,Cluster,AWSMachineDeployment,Machinedeployment) and removed the weird values like uuid self-link creationTimestamp and so on, and only kept the pure CR data. Also removed status field.

Then deleted cluster via gsctl delete cluster and waited until the cluster is gone. Checked the AWS console and saw 3 NAT gateways in deleted state.

I recreated the cluster with the YAML files I dumped before, which created a cluster with the same IDs.

After that checked logs of aws-operator to see that it only found 3 NAT Gateways and that it ignored the deleted ones. Also checked cluster and CloudFormation stacks.

All was green

[calvix]

Also, AFAIK it is more dangerous to create clusters with the same ID twice due to the KMS encryption keys. That could have unintended side effects and we never really did or recommend that.

We do not recommend it but we know customer do it to keep some specific cluster-ID formula, the KMS seems not to be that much problem so far, only the NAT gateways fucked up the CF.