https://pulumi.com logo
d

damp-book-35965

03/28/2019, 5:03 PM
For EKS workers, Can we also look to do updates in a controlled manner ? As of now as I understand it if a new AMI is generated it will show up as a change for all your instance groups..
f

fierce-dinner-20116

03/28/2019, 5:06 PM
Does this also imply that pods aren’t drained/cordoned when updating a node group though?
I guess this is something that I took for granted at least from GKE
d

damp-book-35965

03/28/2019, 5:10 PM
So looks like the autoscaling group handles it by terminating one at a time and bringing one up at a time with the new AMI..atleast looks like that from the activity history
w

white-balloon-205

03/28/2019, 5:16 PM
Yes - the ASG will replace individual nodes one at a time. The issue tracked in https://github.com/pulumi/pulumi-eks/issues/40 is that this process does not interact cleanly with Kuberenetes node drainging, so when a node is replaced, the pods on that node will die suddenly, instead of being migrated off prior to shutdown. This is something that EKS itself has no support for yet unfortunately. We believe we can help solve this problem in
@pulumi/eks
(we've done similar things for ECS) - and plan to do that as part of https://github.com/pulumi/pulumi-eks/issues/40.