https://pulumi.com logo
b

busy-air-16359

07/02/2019, 2:50 PM
When destroying a stack, shouldn’t the Kubernetes provider be destroyed first and then AWS? Right now, the EKS worker nodes are deleted first, and then obviously no k8s operator can cleanup anymore
b

busy-umbrella-36067

07/02/2019, 2:56 PM
The control plane is managed by Amazon, not having worker nodes should’t affect k8s operations
b

busy-air-16359

07/02/2019, 3:04 PM
sure, but a lot of resources are created by kubernetes operators these days
if you kill the operator by removing the worker nodes, there is no way it can clean up its resources
b

busy-umbrella-36067

07/02/2019, 4:03 PM
I hadn’t thought of that, I think you could simulate the behavior you want by adding
dependsOn: this.defaultNodeGroup
here https://github.com/pulumi/pulumi-eks/blob/master/nodejs/eks/cluster.ts#L667
I’d be interested to hear if that manages to delete the K8S Resources deployed with the provider before it deletes the worker asg
b

busy-air-16359

07/02/2019, 4:41 PM
good point, I’ll try adding that
I thought pulumi would figure out the dependency automatically, because the k8s provider is created from the aws/eks provider
that did the trick, thx
👍 1