Stupid RAID question

[Lennart Sorensen]

On Tue, Oct 04, 2011 at 10:43:37AM -0400, Andrej Marjan wrote:
> Thanks to everyone for the informative responses. It seems that among other
> things, I got the terminology wrong: I'm actually looking at using mdadm
> with 2 drives in a RAID 1 setup with drives from different batches or
> manufacturers. I have good backups, so the specific failure case to be
> avoided is when the drive dies and I have no computer until I get another
> drive and waste a bunch of time re-installing. Anything more (RAID5, RAID10)
> seems like it would be overkill for me.
> 
> So I think the solution would be, for each drive:
> - 512MB RAID1 boot partition, formatted to ext3
> - 8 GB swap

Swap must be raid 1 as well.  Might as well put it in the lvm in fact.

> - rest of the drive RAID1 partition with LVM for root and home, all ext4 at
> this time
> - grub in the MBR of both disks

Yep.

> Done right, this is supposed to survive a single hard drive failure (either
> disk can boot into a working system by itself) and give me a few days to
> replace the other drive and rebuild the RAID.
> 
> Am I missing something?

My personal experience has made me never ever want to try root on lvm
ever again.  if something goes wrong with the lvm it is much harder
to recover if you can't boot the system normally until it is fixed.
That is why I use a larger raid 1 for root (including boot), and use lvm
for swap, data, etc.  No reason not to swap on lvm.  I do it all the time.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists