How to Fix a Poorly Planned RAID

Friday, March 30th, 2012 | Backup and Restore Solutions, Technical | Grayson Ronk

 

Using Symantec System Restore (SSR) formerly Backup Exec System Restore (BESR) to fix a poorly configured raid setup

The Problem:

Original server configuration was with 5 (73GB) hard drives into 2 raids (hardware):

  1. 1st raid was a raid 1 = 2 drives mirrored (73GB)
  2. 2nd raid was a raid 5 = 3 drives stripped (145GB)
  3. Then a 6th drive was added to the server (146GB)

Setup in Widows Server 2003 (software):

Raid Bad Practice

This is what a poorly planned raid looks like.

About above image:

  1. “Disk 0” = 1st raid is a raid 1 = 2 drives mirrored (73GB)
  2. “Disk 1” = 2nd raid is a raid 5 = 3 drives stripped (145GB)
  3. “Disk 2” = the 6th drive that was added later (146GB)
  4. (C:) = Operating System – current drive space used 12GB.
  5. (D:) = SQL data – current drive space used 237GB spanned over all three Disks.

The main problem is that the 6th drive has no redundancy (single point of failure).  Another problem would be the performance hit from having a software raid spanning across two different types of hardware raid.

The Fix is to replace the 6 hard drives (4 x 73GB & 1 x 146GB) with 5 drives (5 x 146GB).  The new RAID will be done in hardware and the virtual disk will also be done in hardware.  There will be one partition for the C: drive of 20GB and the rest of the space will be the D: drive.

To accomplish this:

  1. We need to make a full system backup using Symantec System Restore (SSR).  (Time = about 3hrs) NO downtime to the production SQL server.
  2. Test the backup by restoring it to one of the spare servers.  (Time = 5hr) NO downtime to the production SQL server.
  3. After successful test; reconfigure the production SQL server drive setup.  (Time = 10hrs) The production SQL server WILL be down.
    1. Add a differential backup of the system to the full backup made in step 1. using SSR.
    2. Take the production SQL server offline.
    3. Remove old drives and install new drives.
    4. Restore the SSR backup to the production SQL server.
    5. Test and put the server back in production.

Setup in Widows Server 2003 (software) AFTER FIX:

 

Normal Raid

This is what it should look like.

About above image:

  1. “Disk 0” = is now a raid 5 = 4 drives stripped (408GB) Plus 1 global hot spare.
  2. (C:) = Operating System – current drive space used 24GB.
  3. (D:) = SQL data – current drive space used 384GB. (No spanning)

Now we can lose one of any drive without data loss.

 

Tags:

You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.

Page 1 of 11

Leave a Reply