[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

LVM volume portability (was: Re: Good fdisk Practices)


On Aug 24, 2007, at 1:18 PM, David Brodbeck wrote:



On Aug 24, 2007, at 12:13 PM, Ron Johnson wrote:

I read recently on this list that LVM is not portable across CPU
architectures, so that you can't just upgrade your mobo to AMD64 and
retain your /home.
Well, now you've got me curious. If so, this is potentially a serious issue, because most rescue disks are 32-bit. If it's true, then an LVM created on a 64-bit system wouldn't be readable with a 32-bit rescue disk. It also might have implications for things like USB hard disks. (These are getting big enough where it might start to make sense to LVM them -- I have one USB array that's 1.5 TB.)
I happen to have a spare AMD64 system and a couple of spare IA32 systems, all with hot-swap drive bays that take the same sleds, so I may try this and see what happens. 

OK, so here are the results.
I installed RHEL Server 5 AMD64 on a dual Opteron system. (Yeah, I know. But I had the CDs handy, and I didn't have any 64-bit Debian CDs.) I stuck a spare 160 gigabyte hard disk in one of the hot-swap bays and created a single LVM partition covering the entire drive. I then initialized it and created a 100 gigabyte logical volume, which I formatted with ext3fs. Then, for good measure, I created a 100 megabyte file of random bits and calculated its md5 checksum. 

First test: Same hardware, 32-bit kernel:
I rebooted the same system, this time using a RIP Linux 2.5 rescue CD. Ran vgscan, vgchange -ay to activate the volume group, then mounted it. No errors. The md5 checksums matched. 

Second test: 32-bit hardware:
I moved the drive to a dual Xeon system and booted RIP Linux 2.5. Again, no problems. The volume mounted cleanly and the checksums matched.
So, I'm concluding for now that there aren't any portability issues when moving an LVM volume group between IA32 and AMD64 architectures. If someone has tried it and had it *not* work, I'd like to know about it, because this is an issue that could potentially affect me at some point in the future.
Reply to: