azure aks system node pool reconfiguration - does it cause node pool to be recreated?

 We can typically choose between manual and automatic approach to auto-scale our cluster when we create our AKS cluster as you can see from here.

So if we have choosen manual approach, if we were to change it to auto-scale, do we need to re-create the node pool.

Here are my findings 

System node pool

- Can i change the node pool from manual to auto without recreating node pool? Yes you can 

- Can i change VM sku? From the portal no but technically yes that will be an operation to delete existing and recreate a new one. I think someone requested this feature to be added 

https://github.com/Azure/AKS/issues/1574

User node pool

- Can i change the node pool from manual to auto without recreating node pool? Yes you can 

- Can i change VM sku? From the portal no but technically yes that will be an operation to delete existing and recreate a new one 

What if we would like to change the sku of our system pool?

We can create a new system pool with the desired vm sku and then migrate existing pods over. Might have some issues with metric-server pods depending on how much node you have. If you have 1 then probably will run into issue. 

As you can see here, these are my node and pods at the moment. 

Then i run

kubectl cordon aks-agentpool-43776398-vmss000000

kubectl drain aks-agentpool-43776398-vmss000000 --ignore-daemonsets --delete-emptydir-data

Then workload gets migrated over. Cluster continue to functions. 

What if i were to delete the last system node pool? 

In AKS, i would get the following eror.

These are quite common configuration changes but sometimes we just forgot and accidentally re-create the node pool when we tried resizing via terraform. Behind the scene it is deleting and recreating for us and cause possible outage.