How to save omniverse state before stoping backend Virtual Machine

Accelerated Computing Aerial Forum
1

Hi everyone,

I am running the Aerial Omniverse Digital Twin on Azure, where the installation created two VMs:

  • A backend server running the digital twin.
  • A frontend server for the GUI.

To avoid unnecessary Azure charges, I would like to stop and deallocate the VMs when they are not in use. However, before shutting them down, I need to ensure that the current digital twin state is saved properly so that I can restore it when restarting.

My Questions:

  1. What is the recommended way to save the state of the Aerial Omniverse Digital Twin before shutting down the backend server?
  2. Is there a built-in command or API to trigger a state save (e.g., a command like save-state or a configuration option)?
  3. Where is the state data typically stored (e.g., Azure Managed Disks, NFS, or a specific Omniverse storage system)?
  4. After restarting the VM, what steps are needed to restore the saved state?

Any guidance or best practices would be greatly appreciated!

Thanks in advance!

2

I am sorry but we do not actually support AODT on this forum. This is just for Omniverse.

I am moving this post to the correct place
https://forums.developer.nvidia.com/c/accelerated-computing/aerial/674

3

Apologies for posting in the wrong forum… and thatks a lot Richard for moving my question to the correct place.

4

@jroa,
We have not qualified AODT on AWS or other clouds, so we cannot comment on specifics.
In general, when the VM is stopped, then, all the AODT state information is retained. If the VM is complete terminated, then all AODT related state information is destroyed.
Any data generated by AODT and written into the clickhouse database is saved into the storage associated with the VM. You could save it elsewhere (download it, if required). You will have to check with AWS where this storage resides.

5

Thanks @kpasad1. We are using Azure not AWS. Azure is prominently addressed on the AODT user guide and it seems to be fully supported. In any case, I think we will keep the VMs running until we complete our tests, at that time we will dump the clickhouse DB contents and pull that data out of the VM before destroying the VMs. Thanks a lot for your comments/suggestions.