-
Notifications
You must be signed in to change notification settings - Fork 11
Description
Azure does not provide any means of restoring the complete VM state, including extensions, extension 'secret' parameters, Azure Monitor data, RBAC roles for the System Identity, membership in load balancer back-ends, VM Reservations, associated tasks, jobs, Automation workflows, etc...
Similarly, there's no "in-place" rollback capability of any kind -- users are always forced to make an additional copy of a virtual machine to restore its state, which as mentioned above, is guaranteed to lead to data loss in many common scenarios.
There are two kinds of VMs in the cloud: stateless "Cattle" that don't need this feature, and named "Pets" with permanent identities that can benefit from this feature. However, in its current state it is only suitable for Cattle, not Pets, making it virtually useless.
What every cloud admin/operator wants is the equivalent capability of Hyper V "checkpoints" -- the ability to snapshot a VM before a dangerous in-place operation. If the operation goes bad, then the restore must roll back then entire VM, not make a partial copy that may or may not be broken and non-functional.
This feature really needs a simple one-click "restore to snapshot" button that rolls back the VM disks in-place without completely recreating the VM with a new identity. (The VM state should be rolled back also, as if it was an incremental ARM deployment.)
A related feature would be the ability to roll back only some disks (e.g.: roll back the OS disk only, but leave the data disks as-is.)
Before replying to this issue, imagine that you have been asked to apply an in-place Windows OS upgrade to an Active Directory Domain Controller with 100K user accounts for a huge corporation. It is in a load balancer (for LDAP), has dozens of Automation Account jobs associated with it, has a System Managed Identity that's used for RBAC, has an Azure Backup history, and has Alerts wired up to VM Insights to notify a 24/7 service desk for any issues.
Would you be personally confident to use Azure Virtual Machine Restore Points in this scenario to roll back from a bad OS major version update?