AWS: Improve worker pool name

kinvolk / lokomotive

🪦 DISCONTINUED Further Lokomotive development has been discontinued. Lokomotive is a 100% open-source, easy to use and secure Kubernetes distribution from the volks at Kinvolk

https://kinvolk.io/lokomotive-kubernetes/

Apache License 2.0

321 stars 49 forks source link

AWS: Improve worker pool name #705

Open johananl opened 4 years ago

johananl commented 4 years ago

In https://github.com/kinvolk/lokomotive/pull/638/commits/20b7b696a61ba6ec706140feeaae6c25dfb3dfae we reverted a change which prefixed an auto scaling group name with the cluster name. Having the prefix has the following advantages:

It makes it less likely to have name clashes (ASG names must be unique).
It helps keep things DRY in the cluster config (the user doesn't have to prefix the worker pool name with the cluster name).
It makes it easy to see which cluster an ASG belongs to.

The reason for reverting was avoiding breaking upgrades. Maybe we can find a way to make this change in two phases across two releases without breaking things? :thinking:

johananl commented 4 years ago

Another problem:

cluster "aws" {
  ...
  cluster_name     = "johannes-test"

  worker_pool "johannes-test-p1" { ... }
}

The above config results in instances with the following name: johannes-test-johannes-test-p1-worker. I cannot omit the johannes-test- prefix in worker_pool because then the ASG is named p1.

invidian commented 4 years ago

The reason for reverting was avoiding breaking upgrades. Maybe we can find a way to make this change in two phases across two releases without breaking things? thinking

One way would be to add new autoscaling group with new name and then resize old group to 0, but we would have to make sure that the operation is done gracefully. Then, in the next release we can simply remove old group.