AKS resource seems to require vnet_subnet_id

[rberlind]

Terraform Version

Terraform v0.11.7

• provider.azurerm v1.3.3
• provider.tls v1.1.0
• provider.vault v1.1.0

Affected Resource(s)

Please list the resources as a list, for example:

• azurerm_kubernetes_cluster

Terraform Configuration Files

https://github.com/rberlind/tfe-k8s-cluster-aks

Debug Output

Error: Error applying plan:
1 error(s) occurred:

• azurerm_kubernetes_cluster.k8sexample: 1 error(s) occurred:
• azurerm_kubernetes_cluster.k8sexample: containerservice.ManagedClustersClient#CreateOrUpdate: Failure sending request: StatusCode=400 -- Original Error: Code="LinkedInvalidPropertyId" Message="Property id '' at path 'properties.agentPoolProfiles[0].vnetSubnetID' is invalid. Expect fully qualified resource Id that start with '/subscriptions/{subscriptionId}' or '/providers/{resourceProviderNamespace}/'."

Expected Behavior

Since https://www.terraform.io/docs/providers/azurerm/r/kubernetes_cluster.html#vnet_subnet_id says that vnet_subnet_id is optional, the AKS cluster should have created a vnet and subnet or used the default for my resource group.

Actual Behavior

I got error above

Steps to Reproduce

Please list the steps required to reproduce the issue, for example:

1. terraform init
2. terraform apply

[mikestoltz]

I am also seeing this issue - when running a configuration file that had previously worked. Since I only pin azurerm to the minor version 1.3, I thought it may have been introduced with the release of version 1.3.3 - however pinning azurerm to version 1.3.2 did not solve the issue.

[rberlind]

I wonder if it depends on kubernetes_version? I specified 1.7.12. I suspect something is more likely to have changed on the Azure/AKS side than in the Azure Provider itself

[rberlind]

curious how long it takes for an AKS cluster to be created. I added creation of vnet and subnet and then specified the vnet_subnet_id. I see lots of stuff created, but Terraform finally terminated with:

1 error(s) occurred:

• azurerm_kubernetes_cluster.k8sexample: 1 error(s) occurred:
• azurerm_kubernetes_cluster.k8sexample: Long running operation terminated with status 'Failed': Code="ControlPlaneProvisioningInternalError" Message="Internal server error"

[rberlind]

Running apply a second time after that error worked and gave me outputs.

[mikestoltz]

I don't think its related to the k8s version. I didnt change anything in my configuration from when it was working last week and now. I'm using 1.8.7 as the k8s version for what its worth.

[rberlind]

thanks @mikestoltz . I sent an e-mail to one of the HashiCorp engineers who works on the Azure provider to let him know about this issue.

[tombuildsstuff]

👋 hey all

A change got released earlier this week to the Azure API where the Subnet ID is now rejected if it's an empty string, rather than being absent from the request (which is why this is now failing). We added a fix for this issue late Tuesday in #1158 which has since been merged and will be released as a part of v1.4.0 today.

azurerm_kubernetes_cluster.k8sexample: Long running operation terminated with status 'Failed': Code="ControlPlaneProvisioningInternalError" Message="Internal server error"

@rberlind from my understanding, this is a capacity issue within the Azure region you're using - from my understanding AKS is still resource strained in particular regions; can you try reprovisioning into another region?

Given this issue's been fixes in #1158 and will be released shortly - I'm going to close this issue for the moment.

Thanks!

[tombuildsstuff]

hey @rberlind @mikestoltz

Just to let you know that v1.4.0 of the Azure Provider has been released which includes a fix for this :)

Thanks!

[rberlind]

Thanks @tombuildsstuff : My code did not include net_subnet_id at all in my original code (which has changed since I posted this). I had:

  agent_pool_profile {
    name       = "${var.agent_pool_name}"
    count      =  "${var.agent_count}"
    os_type    = "${var.os_type}"
    os_disk_size_gb = "${var.os_disk_size}"
    vm_size    = "${var.vm_size}"
  }

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you feel this issue should be reopened, we encourage creating a new issue linking back to this one for added context. If you feel I made an error 🤖 🙉 , please reach out to my human friends 👉 hashibot-feedback@hashicorp.com. Thanks!