1

I don't understand why a separate resource group is created for all of the infrastructure associated with an ACS cluster, and not the Resource Group I specify when creating the cluster? This leave my defined Resource Group with one lonely entity (the ACS Cluster definition) and a whole new Resource Group whose name I don't control. Not a fan of this.

I am currently using the Azure CLI to create my ACS cluster, so I'm "guessing" if I went the ARM route I'd have more control. Still, where does this limitation reside and why?

Here's my CLI command:

az acs create -n=int-madraskube -g=internal-acs
--orchestrator-type=kubernetes  --agent-count=2 --generate-ssh-keys --windows --admin-username={myadmin} --admin-password={mypassword} --service-principal={sp_guid} --client-secret={secret_guid}

And I end up with two resource groups:

internal-acs

internal-acs_int-madraskube_westus2

Community
  • 1
  • 1
BrettRobi
  • 3,793
  • 7
  • 40
  • 64
  • What version of acs-engine are you using? Do you create the resource group before executing the cluster creation? – radu-matei Sep 14 '17 at 08:17
  • Hmm, not using acs-engine directly. I'm using the Azure CLI to create everything. I did create the resource group manually first. – BrettRobi Sep 14 '17 at 17:48

1 Answers1

2

This is the new design of the ACS (v2) in some selected regions. In the past (the v1), the created resources are in the same resource group as container service resource is. It makes it really hard to clean up those resources when you delete the container service resource. With those being in a new resource group (preferably locked down and invisible, but not possible right now), it's one step closer to managed service.

For more info about ACS v2, https://github.com/Azure/ACS

Weinong Wang
  • 146
  • 3
  • So interestingly I just created a new cluster and it used the RG I specified and put everything in that (no second RG). Just the way I expected and wanted. The only difference being it's two weeks later than the first cluster creation and it's eastus2 vs westus2. Does this mean I now have an "older" cluster design? I can see the benefits of using a second RG if you were truly going to hide all of that from me, as in a more truly PaaS system, but this in-between state is confusing and not ideal. – BrettRobi Sep 14 '17 at 17:52
  • yep, what you experienced is the v1. See the regions of v1 and v2 RP at https://github.com/Azure/ACS/blob/master/kubernetes-version-support.md – Weinong Wang Sep 14 '17 at 20:12