Fix a broken RAID5 1

[DedeMole]

I have a V880 server running solaris8, with 12 36Gb hardrives... We just recieved all new 72 gb hardrives. My assistant thought that if he just brought down the server and put in the new disks everything would work (he was very wrong).... So then in a panic he just replace the OS harddrive and tried to bootup the machine with the new HD (which doesn't work either because 1/2 of the new drives haven't been formated yet).... Anyway....

I have now replaced all the drives AND in the correct order they were orginally in but my server won't startup, and I think it's because the RAID is corrupted now....

How can I restart/reinitialize the old RAID5 I had working before his help???

 

[KenCunningham]

Who's your assistant? Igor from Young Frankenstein springs to mind!

Do you have a backup? If so, it would possibly be easier to reinstall Solaris, do the necessary RAIDing then restore your data. Other than that, could you boot from CD then investigate what and where the disks now think they are?

 

[coffeysm]

Question number one what RAID software are you using? Question number two did you have your boot drives mirrored?
Question number three you moved this person to the Windows team right?

If you have a backup though Ken's suggestion might be the easiest. Booting off cd might or might not help you in this situation.

 

[DedeMole]

Um thanks KenCunningham and coffeysm.... And yes he is back to the windows team, unfortuantaly he was suppose to have been my UNIX expert and I barely know how to setup a RAID5 myself. I'm not sure what software was used to setup the orginial disks.

OK-- the boot drive was not mirrored. If I start over with building a new OS and RAID (ughgghgh that will take time) do I have to make sure all the drives that will be used for the RAID be formated beforehand and if so how should I partition those drives???

If not then when I build the new OS disk what should I tell the system to do with the 11 other drives??

In the meantime I will try booting from the CD and investigate the current state of the old RAID5.

 

[coffeysm]

The software used can pretty much make or break you in this situation, most likely it is Veritas or Disksuite (SVM now). The first thing you need to do though is get the OS up and running.

1. boot cdrom -s
2. fsck -y /dev/dsk/your-boot-disk-slices 
3. mount /dev/dsk/boot-disk0 /mnt
4. check /mnt/etc/vfstab and look for your RAID5 mount comment them out
If it appears as: /dev/md/dsk - DiskSuite (SVM)
If it appears as: /dev/vx/dsk - Veritas
5. check /mnt/etc/system
Veritas will have entries like (Comment them out with a *)
forceload: drv/vxdmp
forceload: drv/vxio
forceload: drv/vxspec
forceload: drv/dad
Check /mnt/etc/rc?.d directories for S??vxvm-blah-blah scripts and make them little s??vxvm-blah-blah

DiskSuite will have entries like (Comment them out with a *)
forceload: misc/md_trans
forceload: misc/md_raid
forceload: misc/md_hotspares
forceload: misc/md_sp
forceload: misc/md_stripe
forceload: misc/md_mirror
set md:mddb_blah_blah_blah

They might have changed the script names best bet is to go into the rc?.d directories and do a grep meta * and change the names of those files also.

6. Try booting up now to get the OS up

 

[coffeysm]

To answer your other question yes restarting from scrath will take a LONG time. You have to install the OS, the RAID software, initialize the RAID5 which can take hours at a time, and finally restore the data which can also take hours.

You just tell Solaris to ignore the 11 other disks you have available. No you will not need to format the drives since the RAID software will do all that for you when it is initializing.

 

[coffeysm]

Forgot to mention though if you are indeed using Veritas, starting from scratch can be sometimes easier. As long as you have the drives in order, you can just install the OS and Veritas. There are a few commands you ran after that will allow you to reimport your group and data without initializing or restoring from backup. I forgot how to do that off the top of my head though, been a pretty long time.

 

[coffeysm]

Thanks for the start, but you might want to give it to me after it works or not....lol