Makes sense to keep a spare USB boot drive plugged in?

[dfosbenner]

I presently use 16GB USB thumb drives on my (4) ESXi hosts.  I am wondering if it makes sense to have a second USB drive attached, in the event of a failure on the primary.  Does anyone else do this?  When I do server maintenance, it's usually always in the evening via remote access.  If I had a USB drive go down, that would be when I'd discover it.  A second USB would give me something as a fall back.  Just wondering what the consensus is on this.

[dabson]

A second USB boot drive is a novel idea. I'd probably ask myself a few questions before deciding on whether it fit any needs. You didn't mention which version of vSphere you're using, so I'll cast a wide net.

• Am I confident that the failure of the primary boot drive can and will be detected so that it can addressed?
  • If not, is it acceptable to not know if a host has failed over to the secondary drive?
• Do I have HA enabled on my (presumably clustered) ESXi hosts?
  • If yes, can resolution of the failed host wait until next business day?
• Can two USB drives physically coexist connected to the same server?
  • Sounds silly on the surface, but depending on USB port placement and USB drive shape, you may encounter issues.
• Is there an alternative available during a USB drive failure, for example vSphere Auto Deploy?
  • Mind you, if you have vSphere Auto Deploy, why not consider it instead of USB drives?

I'll be curious to see if anyone else chimes in that is actually doing this in the wild since, as I mentioned, it does seem a novel idea.

Please make sure to come back and let us know if you decide to roll out secondary USB drives and how it goes.

[dfosbenner]

I am using ESXi 5.5.  Here's the scenario I had in my head:

I connect remotely via DRAC (Dell remote access), or HP ILO.  I patch ESXi, reboot it, and the server doesn't come back up.  Maybe it's a PSOD.  Maybe worse.  From what I've read, these little USB drives do fail.  I keep a bootable ESXi install CD in the drive of my hosts.  Using my DRAC, I can specify to boot from the CD, and then do a new install & patch of ESXi to the second USB drive.  Takes about 30 min, bring it up, either restore my config or re-add my VMs, and I'm back online.

That was the idea, anyway.  I don't have clusters, HA, etc.

[dabson]

Conceptually, as long as you're comfortable that your system will always boot from the primary, you could even consider making your secondary USB drive a clone of the primary. You'd have to ignore the VMFS partition from this drive, assuming you provision one, during normal operation. In this case then if your primary fails to boot then potentially your server may be able to automatically boot from the second.

If it's not a full USB drive failure and, such as you suggest, you encounter a PSOD or similar, then this automatic fail over won't occur of course. You'd still have to manipulate the boot order. For the situation you're describing a secondary clone may do the trick and save you from going through the install process. Just boot to secondary, patch and continue.

Assuming this works then maybe you add a step to your regular system maintenance to ensure the secondary drive is cloned from the primary (or installed and patched from DRAC) on a reasonable schedule.

Still, this approach holds some promise. I'd suggest giving it a try.

[JarryG]

I have something similar, albeit not USB-sticks, but CF-cards (with CF-to-SATA adapters): 2x CF, each with ESXi installed (but not in raid). It serves me as backup for the case when a) one CF gets damaged, or b) I screw something up while updating/patching ESXi. And using KVM remote-access I can reset server, go to bios-settings and switch boot-disk from the 1st CF to the 2nd one, if necessary...