If a PCIe HBA added to a client vm and both possible host HOSTS have identical hardware setup including the same model HBA can HA be kept?

C

Citizen

New Member
Jan 11, 2024
2
0
1
If a PCIe HBA added to a client vm and both possible host HOSTS have identical hardware setup including the same model HBA can HA be kept?

In short I have a disk shelf connected to the host with a HBA. I do want to keep it as High availability as failover. So that if PVE1 fails the VM will move to PVE2 and mount the HBA in that other node. Effectively only having a possible reboot down time and live access to data. This would allow only the Root drive to be on the host storage with a replicate job to both nodes (would have it replicate to the others too but they would not have access to the HBA so would be just backups and not able to work)

All info i see if if i pass them to vm the replicate is broken. As both hosts would have the same config and same model HBA with same PCIe mount point they effectively both have access to the storage allowing this to be a poor mans high availably cluster with only 2 nodes.

My question is as my setup would be identical setup on both vms would this work? Is their a way to get this to work?
 
While I do NOT use anything of this I want to add my 2 €¢:

When you PCI-passthrough a hardware device then the host/PVE will never touch it. The guest alone is in charge of managing that device. That's the whole point of this mechanism.

Inside a PCI device there is (possibly a lot of) data, building a "state". Only the guest knows about it.

When PVE would (hypothetically) migrate the VM to another node with the exact same hardware available, how would it possible to migrate that internal state?

For the host there is no way to do that.

---
Edit: my text is obviously about live migration. Stop-Mode-Migration should now be possible, see #4 :-)
 
Last edited:
You must log in or register to reply here.
Top Bottom