Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations strongm on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

recovering lost raid 5 data

Status
Not open for further replies.

clarkdo

MIS
Jun 27, 2002
13
US
I have 4 identical servers - each has three 18.2GB drives in a raid5. One of the servers has failed - data totally wiped out. Can I rebuild that server by putting one drive from the failed server into one of the good servers and letting it rebuild and then repeat for the second drive and then repeat for the third?
 
First, if the data is wiped out, what will it be rebuilt from? Tell us more about the failure & what you're trying to accomplish. On the failed server, did one or more drives fail, the raid controller, or something else in the box? What makes you think that the data is totally wiped out?
 
The server tried to reboot last night as scheduled but the message this morning was that the smart array was not configured. Pressing F1 to continue tells me that it is a non-system disk and to replace with a bootable disk.

I attempted to repair the windows 2000 installation but the message is displayed that it can't find a windows 2000 installation to repair. When I attempt to reinstall windows 2000 (I was going to install to a new directory to keep the old data intact), I am told that I must create a partition as the entire drive is unpartitioned.

I have 4 identical servers. My thought is to swap disk 1 from the bad server with disk 1 in a good server and let the disk rebuild. After it rebuilds, I would swap disk 2 from the bad server with disk 2 in the good server, and then finally the same for the third disk. I would then have cloned the three disks from a good server and could use them in the bad server.

Any chance that it would work or just a waste of time?

 
This will not work as the data on these drives would be identical and that is not what is contained on a set of 3 drives. In a set each drive has different informaton.

You would also loose all chances of retreiving any of the data. If you are at this point, leave the drives in the failed server and reformat them and rebuild the array. Then reinstall the OS. I am somewhat certian that a WIN server OS cannot be installed to a RAID 5 array, but if that is how you were configured then what ever you did must be possible,

rvnguy
"I know everything..I just can't remember it all
 
In order for a RAID5 array with 3 drives to fail, you must have lost at least 2 drives. RAID5 can tolerate a loss of 1 drive in an array, but not more than 1.

In order to rebuild the array, all member drives, less the defunct (bad) drive, must be present. Using one drive at a time won't work as there simply isn't enough parity information on any one drive to rebuild the entire array.

Is there any type of RAID management built into the RAID controller BIOS? Usually there is so that diagnosis and repair of a failing array can be made.

And as a preventive measure, does your RAID controller have any RAID monitoring software that can be added to your Windows servers? I ask as you have 3 others servers running RAID5. You should have some type of RAID health monitoring software so that if 1 drive fails, you can take action to replace it before another drive goes bad.

And a Windows OS can be installed on a RAID5 array as long as the controller's driver is fed to the OS via the F6 option very early on in the non-GUI installation phase.
 
1 - The 4 servers have the same hardware, software and OS (W2000 server) - three 18.2 disks in each.
2 - Server 1,2, and 3 are OK, server 4 has failed. I do not need anything from server 4 as it is the same as 1,2,& 3. There is no data on this server - only apps. Data is saved in another location.
3 - My thought was to clone server 1 to server 4 by having the raid rebuild the disks one at a time. Remove disk 1 in server 1 (leaving 2 good disks) and replace it with the erased disk from server 4. When it rebuilds, I would have cloned disk 1. Then I would do the same with disk 2. And then with disk 3.

Thanks for responding.
 
3 - My thought was to clone server 1 to server 4 by having the raid rebuild the disks one at a time. Remove disk 1 in server 1 (leaving 2 good disks) and replace it with the erased disk from server 4. When it rebuilds, I would have cloned disk 1. Then I would do the same with disk 2. And then with disk 3.

This will not be successful...as stated these need to be part of a set and not three identical images re-built from another server.

Each member of a RAID 5 array contains unique data and is part of the set.

rvnguy
"I know everything..I just can't remember it all
 
Interesting approach, rebuilding each striped disk, and one I've never heard attempted before. I personally don't think it'll work, but don't let that stop you.

My gut tells me that the RAID controller in the broken server won't recognize the RAID5 volume once its rebuilt, but you never know. If your theory works, then you should only have to rebuild two of the drives then rebuild the third from them.

If it somehow does work, just be aware that you will then have two servers with the same name, IP, SID, access rights, etc.
 
I missed the d1 to 1, d2 to 2...

This might work but I tend to agree with Freestone in that the controller serializes the disks and this will most likely not match what is recorded in the failed servers contoller BIOS(not the system BIOS).

If you a determined to try this, rebuilding takes a good amount of time and the server needs to be down to prevent changes to the data to insure a match.

I would shut the sever to be cloned down and in another PC clone each drive with Acronis, Ghost or another that you like. Keep them marked carefully and place each on the same channel as on the donor.

As Freestone noted you will then have to rename, & reset several things.

Another caveat might be that the HAL will allow 3 points to be different without activation. As most servers have two nic's and a nic counts as three you will have a hardware difference = 6 and it will balk on boot due to this. Unless Server 2K3 is different in this respect from XP

rvnguy
"I know everything..I just can't remember it all
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top