I would like to keep each VM in separate resource group for ease for lifecycle management. I have a cluster containing n VLMs.
So I create one resource group for common things like public IP, load balancer and put availabilitySet declaration into it because is also must be shared between VMs.
Then I create VM in separate resource group and reference to availabilitySet with
"availabilitySet": {
"id": "[resourceId('Microsoft.Compute/availabilitySets',variables('availabilitySetName'))]"
},
Of cause 'availabilitySetName' is defined.
When I deploy my template I get an error saying
{"error":{"code":"BadRequest","message":"Entity resourceGroupName in resource reference id /subscriptions/a719381f-1fa0-4b06-8e29-ad6ea7d3c90b/resourceGroups/TB_PIP_OPSRV_UAT/providers/Microsoft.Compute/availabilitySets/tb_avlbs_opsrv_uat is invalid."}}
I double checked that resource and availability set name are specified correctly.
Does it mean that I can't put a set in separate resource group from VM?
Unfortunately, having a VM use an availabilitySet in a different resource group is not supported :(.
First of all, let me ask you why you want different resource groups? I strongly believe that you're overthinking it with multiple resource groups. A resource group is basically your "Entire System" and within the boundaries of one solution, you should only have one resource group for production, one for beta/staging etc, but never mix.
If you're selling SaaS to your customers it would make sense to have one resource group for each of your customers.
And as you know, a Resource group is simply a way for you to link together and manage all of the assets in your solution; vm's, storage, databases etc under one common name. I am very doubtful as to why one would want to consider multiple resource groups in a single solution, however, I am always willing to learn :)
Availability groups
Now, Availability groups are a different thing. This has to do with "Update Domains" and "Fault Domains" for your VM instances. Because Azure does not keep 3 separate VM's for you, as it does with most of it's PaaS services, you have to manage these yourself to ensure full uptime. Basically, when you're adding two or more VM's in an Availability Set, you're ensured that planned or unplanned events, at least one of the VM's will be available to meet the SLA.
Trying to combine the two in an effort of preventing downtime may sound like a good idea, but it is not solving any problems that I'm aware of. Like the old saying goes: if it aint broke, don't fix it :)
