How to share a SCSI device from one host to another

[eldertree]

Hello,

I have 2 hosts ESXi01 & 02 running on VMware 5.5.0. There is a tape drive IBM TS2250 connected to ESXi02 and a virtual machine on ESXi01 which needs access to this tape drive. When I try to add it in VM settings it's grey and unavailable, but it's available for VMs on ESXi02 where it's connected.

Funny thing is that it worked before but after power failure, I had to start the VM quickly without any physical access to the server room, so I had to temporarily delete this SCSI device (which was off) to be able to start the VM and I can't add it back now. I'd be grateful for any tips.

[daphnissov]

Passing through tape drives to VMs isn't supported by VMware to begin with and, as you noted, doesn't quite work being a SCSI passthrough. The recommendation is to have a physical server mounting that tape drive. If other VMs need access to that tape drive, they need to somehow go through the physical machine acting as a gateway.

[eldertree]

Ok, but it worked fine for a few years and I did hundreds of backups using this tape drive until this unfortunate event. Maybe I just have to restart my hosts with tape drive on? After a power failure, the tape drive had to be powered on manually, so when my hosts were starting it was off. Maybe that's the case?

[daphnissov]

I don't know how you were ever able to present a tape drive mounted to one ESXi host to VMs on a totally different ESXi host unless the VM, which is now on a separate host, was at one point on the host with the tape drive. This is one of the reasons you don't do that to begin with.

[eldertree]

So, it is possible that at the beginning the VM was on another host and previous admin migrated it. SCSI device was working until I deleted it and there is no way to make it work again other than connecting the tape drive to the current host or migrating the VM?

[daphnissov · Accepted solution]

there is no way to make it work again other than connecting the tape drive to the current host or migrating the VM?

That's right. Or, you know, use this experience and correct your environment so you're not doing that to begin with, and inform management that this should have never been done from day one.

[eldertree]

Migration to another host solved the problem. Thank you