[AZURE] Frequent FreeBSD 11.4 OSProvisioningInternalError

[Annie Latsko]

Hi,
My team is using the latest Marketplace FreeBSD 11.4 image (thefreebsdfoundation:freebsd-11_4:11_4-release:latest) to deploy VMs. Our deployments are frequently failing due to OSProvisioningInternalErrors. There is little information provided in the error itself, or in the deployment’s respective ARM logs.
Deployment failed. Correlation ID: e2347a6f-f65c-400f-a8f6-af6cf10337d3. {
  "status": "Failed",
  "error": {
    "code": "ResourceDeploymentFailure",
    "message": "The resource operation completed with terminal provisioning state 'Failed'.",
    "details": [
      {
        "code": "OSProvisioningInternalError",
        "message": "OS Provisioning failed for VM 'tags-v4-5' due to an internal error."
      }
    ]
  }
}
We see this when we try to deploy v3 and v4 VMs, in both eastus and eastus2.
When this error occurs, the VM will still be in a Running state. I can connect to it using the serial console. One thing that I noticed is that when I run a `camcontrol devlist`, I don’t see scbus3.
?root at freebsd:~ # sudo camcontrol devlist
<Msft Virtual CD/ROM 1.0>          at scbus1 target 0 lun 0 (cd0,pass0)
<Msft Virtual Disk 1.0>            at scbus2 target 0 lun 0 (da0,pass1)
root at freebsd:~ # mount
/dev/gpt/rootfs on / (ufs, local, soft-updates)
devfs on /dev (devfs, local, multilabel)
On a healthy VM, scbus3 is there:
?root at freebsd$ sudo camcontrol devlist
<Msft Virtual CD/ROM 1.0>          at scbus1 target 0 lun 0 (cd0,pass0)
<Msft Virtual Disk 1.0>            at scbus2 target 0 lun 0 (da0,pass1)
<Msft Virtual Disk 1.0>            at scbus3 target 1 lun 0 (da1,pass2)
?root at freebsd ~]$ mount
/dev/gpt/rootfs on / (ufs, local, soft-updates)
devfs on /dev (devfs, local, multilabel)
/dev/da1p1 on /mnt/resource (ufs, local, soft-updates)

We have deployed VMs extensively using the 11.3 image (not maintained by the FreeBSD Foundation) and did not see that error. Anecdotally (we don't have hard data), we also weren't seeing it until December. We have VMs from when we were testing with fewer issues before December and they're on 11.4-RELEASE-p5, as opposed to 11.4-RELEASE-p8 now.
Has anyone else experienced this?