Planning a cluster with HA vm with USB modem.

[kwinz]

I have successfully evaluated Proxmox in a HA cluster. And now I just ordered new servers from Dell that this will be running on commercially.

One thing where I am not sure yet if it's possible: One of our current machines is using a USB modem to query an embedded legacy machine every night. It's an important task for us internally so besides the compute and store that the new Proxmox cluster will do I would like to also move this querying into a VM on the new Proxmox HA cluster.

If I now connect the same USB modems https://www.usr.com/products/56k-dialup-modem/usr5637/ on all servers of this VM's HA group.
Can I pass this USB device to a Virtual Machine, and when the server dies for whatever reason the VM gets migrated to another server it will pick up the USB modem from the different server?

This old thread makes me think no but my case is a bit special:

Keep in mind vendor and device ID would stay the exact same since all servers have the same modem.
And the USB modem has no state (after the connection is done which takes 5min each time) the modem hangs up and is idle the rest of the day.
And finally I don't want to share a single USB modem over a network solution. By giving each server its own separate identical modem that takes the modem out of the single point of failures as well.

Thanks in advance!

 

[t.lamprecht]

Hi,

Can I pass this USB device to a Virtual Machine, and when the server dies for whatever reason the VM gets migrated to another server it will pick up the USB modem from the different server?

If it is really the same, i.e., matches the same "pass this device through" config line for all nodes it should run it could work.

You won't be able to live migrate it, IIRC, but recovery on node failure is not a migration at all, it involves "stealing" the VM config after the node it was on got (self)fenced and the HA stack got its lock. Then it's started again on a other VM, and that can work.

You can simply evaluate it by trying a shutdown + move + start of that VM, if that works out you should be good for HA too.
Note, restrict that VM to only those nodes it can run on at all via HA-Groups, tick on "nofailback" and "restricted" to avoid any movement of that outside those, even if no other node is available (as it cannot run there anyway).