I have two identical tape drives, one on each server. One is acting weird and I wanted to see if it was a driver problem or the tape drive itself by swapping them.

However, I don't want to shut down the servers. Is this possible? If so, what is the procedure (do I need to unmount first and then remount after the swap for example)?

As a rule, no. SCSI usually doesn't react well to being unplugged while hot, unless it's specifically noted as a hot-pluggable device. As a rule, tape devices usually aren't. You don't say anything about the hardware, so it's hard to say. Check the manuals that came with the devices and see if they're hot-pluggable.

Sorry, they are DLT-V4 drives by Quantum.

Doesn't react well means...? I could break something, or it just won't come back up again until I reboot.

It's not going to damage the hardware, but quite possibly will not operate until after a reboot.

The one thing that you have going for you is that the drives are identical, which means that the same driver is required for both AND the servers probably give them the same device name. THAT means that in your specific scenario, it's possible that they will work if you swap them because the device is already mounted and the driver is loaded.

However, as previous poster noted... as a general rule of thumb, you must reboot before SCSI devices will function because the nature of SCSI is comparable to a physical hardware connection, such as a PCI riser card, or a stick of RAM. That's why the channels have to be terminated.

I see.

One system is Solaris 5.9 and the drive is named: rst0 and various numbers. The other one is FreeBSD 5.2.1 and the drive is named sa0.

Well, I suppose I will schedule a reboot of the server.

So what about the SCSI hard drives on this system? It is an HP Proliant ML350 G3.

The drives are accessed by opening the front panel open (it's on a hinge like a door) and just pulling the drives out. I expect that I can at least unmount those and pull them in and out if necessary.

Again, check your manuals. Some systems have hot-plug drives, some don't. And Linville79 hit it right on the head...it won't break anything, but the device may give you flaky results, until you reboot.

As a rule, whenever I dealt with SCSI devices, I would power everything off totally. If I had an external device, I'd power it off too, and power it back up first, and let it go through whatever self-checks it had to, before I powered the host back up. Got much better results that way. Also, if you just plug in or unplug a SCSI drive, the system may decide it's not happy anymore, and crash. Results go all over the map if you don't have hot-plug devices, so I'd always err on the side of caution.

Hmm, well, it's got a hot swap cage on the front:

http://www.open-mag.com/9973483279.shtml

And the drives are all HP drives purchased from HP, so they probably are hot swappable.

EDIT:

Meanwhile, I wonder if rebooting without turning off the power of the external tape drive is what's causing my strange tape errors.

This shouldn't be a problem, because generally, you'd need to have the external SCSI device powered up and fully booted before powering on the server, just as TBone said in his first comment.

The proper procedure would be to have everything powered off... then the tape drive would be powered on first... after it completed its POST then you'd power on the server.

By just rebooting the server, you're not really doing anything out of order, just skipping the power off/power on again steps for the tape drive. It's still in the same state that it should be in before powering the server on at that point.

That makes sense.

I went ahead and used the tape drive as is. The strange garbage bits don't seem to be affecting the integrity of the files, even the last files in the archives I put on the tape, so I guess it's not a huge issue.