Blob (vhd file) not being deleted?

[vmary2014]

Hi Not sure if this is an option available or not but here is: When I create a VM on Azure, then destroy it, it appears the VHD file (127GB each time) remains in the container. Why is the VHD file not being destroyed? Is it possible to get it removed? Thanks Vincent

[matt-richardson]

I'm also facing this issue. Any news, team?

[devigned]

Is this still the case? Could you share your Vagrantfile (minus the sensitive info)?

[matt-richardson]

Yep, happens with 1.3.0:

Vagrant.configure(2) do |config|

  hostname = "vagrantdsc.local"
  config.vm.guest = :windows
  config.vm.communicator = "winrm"

  config.vm.synced_folder ".", "/vagrant", disabled: true

  if Vagrant.has_plugin?("vagrant-multi-hostsupdater")
    config.multihostsupdater.aliases = {ip_address => [hostname]}
  end

  config.vm.provider :azure do |azure, override|
    azure.mgmt_certificate = 'azure-management-cert.pem'
    azure.mgmt_endpoint = 'https://management.core.windows.net'
    azure.subscription_id = ENV['AZURE_SUBSCRIPTION_ID']
    azure.vm_image = 'fb83b3509582419d99629ce476bcb5c8__SqlServer-2016-RTM-13.0.1601.5-Web-ENU-WS2012R2-CY16-SU03v5'
    azure.vm_name = 'MyVM'
    azure.cloud_service_name = 'MyCloudService'
    azure.vm_password = ENV['AZURE_VM_PASSWORD']
    azure.storage_acct_name = 'my_storage_account'
    azure.vm_user = 'my_admin'
    azure.vm_size = 'Standard_DS2_v2'
    azure.vm_location = 'Australia East'
    azure.state_read_timeout = 600 # seconds
    azure.winrm_transport = [ 'https' ]
    azure.tcp_endpoints = '3389:53389' 
    override.vm.box = "dummy"
    override.vm.box_url = "azure.dummy.box"
  end
end

[devigned]

Could a reasonable response to this be, please use 2.0.0.pre2 or greater?

[matt-richardson]

Unfortunately, not for my scenario, as I need to test against both ASM and ARM modes (its for an Azure VM extension).

[matt-richardson]

It looks like its not releasing the lease on the disk - not sure why

[devigned]

Last I investigated this, that is exactly what I found @matt-richardson. The lease on the disk took too long to release and we give up trying to delete it.

I honestly didn't dig too deeply into it, but rather used a naming convention and a automated script to clean up after myself.

[matt-richardson]

Thats the approach I took:

disk_name=`azure vm disk list VM_NAME --json | grep -i "mediaLink" | sed 's/.*vhds\/\(.*\)\",/\1/'`

echo "Running 'vagrant destroy -f'"
vagrant destroy -f

echo "Cleaning up orphaned disk" # for some reason, its not releasing the lease, so its not deleting the disk
azure storage blob lease break --blob $disk_name --container vhds --account-name $AZURE_STORAGE_ACCOUNT_NAME --account-key $AZURE_STORAGE_ACCOUNT_KEY --duration 0
azure storage blob delete --blob $disk_name --container vhds --account-name $AZURE_STORAGE_ACCOUNT_NAME --account-key $AZURE_STORAGE_ACCOUNT_KEY --delete-snapshots include --quiet

It'd be better if we didn't have to do that though. I didn't notice for ages, then by chance, I found 25 x 127GB files in a storage account that I wasn't expecting, so I can see this impacting others.

[devigned]

I think this could be added pretty simply through https://github.com/Azure/azure-storage-ruby/blob/master/lib/azure/storage/blob/blob.rb#L571.