Koozali.org: home of the SME Server

RAID1 Restore Problem

Willy Roesen

RAID1 Restore Problem
« on: February 20, 2003, 12:42:19 PM »
I came up on a curious problem the other day.
My hda ceased breathing suddenly, and had to be replaced.
I followed the restore howto for 4.1.2, since the commands seemed to work, even though my version is the 5.6, and the RAID rebuild went fine, btw. But then the fun started....
First problem: The system still wouldn't boot without the diskette.
Second problem: The new hda never came online.
The RAID simply discarded it and continued in degraded mode on hdc.

Luckily I have a tape back-up station, so I managed with a fresh install and subsequent restore.
... But that is not the right way to do it, or .....?

PS! Thank you very much for the fsck hints in this list !!! I really tried everything to make contact with the 'dead' hda.

Greetings
Willy

John Crisp

Re: RAID1 Restore Problem
« Reply #1 on: February 20, 2003, 03:19:10 PM »
Willy, I have a similar sort of problem. First stop is at www.myezserver.com and read the raid recovery howto (I think you can also find it in the howtos here)

That will tell you how to rebuild the array.

I have a had a similar problem in that the machine will not boot with the first disk removed which indicates that lilo has not been written correctly to the second disk. Don;t the answer at the minute - have a search through the forums for more info.

Good luck

B. Rgds
John

Willy Roesen

Re: RAID1 Restore Problem
« Reply #2 on: February 20, 2003, 05:15:39 PM »
Thank you for pointing me in Darrells direction.
As nearly always when you search there, a meaningful answer came up.
I hope I'll never have to utilize it, but IF .... I have archived the howto with my server-documentation.

PS! I don't think Lilo is supposed to make the second disk bootable. I remember having read something about it, but not where, I'm afraid.

Greetings
WyRon

Nathan Fowler

Re: RAID1 Restore Problem
« Reply #3 on: February 20, 2003, 06:09:50 PM »
Willy, I run RAID-1 on SME 4.1.2, and you're right, out of the box lilo will only boot the first disk.  This isn't as "fault tolerant" as one would hope.  I was able to make the second disk bootable by changing the /etc/lilo.conf line to point to /dev/hda and then I re-ran lilo.  I then changed it from /dev/hda to /dev/md0 or whatever my boot partition was at the time, it's been so long ago.  To make a long story short, both disks are bootable, so if one dies I can immediately start using the second, even if a reboot is required.

Willy Roesen

Re: RAID1 Restore Problem
« Reply #4 on: February 20, 2003, 06:31:43 PM »
Nathan, thank you.
That is probably the subject of which I remembered only a fragment.
I hope the same procedure is valid in 5.*, so John Crisp won't have to look any further.
But (for a morale) always rememer to have a boot-diskette at hand (and some documentation, if you are a newbie (as I certainly am)). You never know !

Brgd's/Grtg's
WyRon

John Crisp

Re: RAID1 Restore Problem
« Reply #5 on: February 23, 2003, 11:55:49 PM »
Nathan & Willy,

I think that this has finally got me round the situation - I have read and read and couldn't find the answer as to why the second disk wouldn't boot - at last I know. Will have a go with Nathans answer, but it now everything makes perfect sense.

Just a shame that 1. there is not any obvious documentation that I could see too that effect and 2. for a server offering mirrored drives - RAID, it should be able to do it out of the box.

A comment in the E-smith docs wouldn't have gone amiss.

Anyway, many thanks.

B. Rgds
John