I got a great note from one of our system engineers (Mark Jordan, a terrific SE who does a great job) this week concerning the use of BMR (BareMetal Restore/Recovery, sometimes called Bare Metal Restore/Recovery) that I think is pretty interesting. Most of the focus on Bare Metal Restore these days is on P2V (Physical to Virtual) and V2P (Virtual to Physical), but Bare Metal in a P2P (Physical to Physical) mode – even in the simplest case of similar Bare Metal (versus dissimilar Bare Metal, where you’re backing up one physical computer and restoring to another) can get interesting with all of the heterogeneous storage options out there (DAS, NAS, and SAN.)
So with no further ado, here’s the story.
Bare Metal (or BareMetal) Backup and Restore: Mark Jordan’s Story
I thought you would interested in two ways I used bare metal restore/recovery this week at a new customer site. We needed to increase the size of the C-volume on two Windows Server 2003 systems, one the Exchange server.
In case one, the customer had a single array holding two partitions and volumes. The C-volume was 12GBs with the balance of the array (more than 300GB) an unused D-volume. After making sure we had a good BareMetal backup (not needed assuming there were no problems with the volume expansion), we used the Windows storage manager to delete the D-volume and the second partition. We then booted to the Windows PE BMR CD and used the diskpart tool to expand the C-volume. The series of commands used were:
select disk 0
select volume 1
expand volume size=38912
This series of commands expanded the 12GB C-volume to 50GB. The size value is presented in megabytes. After performing these steps we booted back into Windows and partitioned the remaining unallocated space on the array and created a new D-volume. If there had been data on the D-volume, we could have then performed a restore of a file backup to bring back that data. This procedure took very little down time with the bulk of the few minutes of downtime being the time it takes to boot the server.
The second case was a little more complex. We had a Windows Server 2003 system running Exchange 2003. We planned for 90 minutes of downtime after regular work hour to complete the expansion of the C-volume. In this case we had five volumes created on five different LUNS on a Compellent SAN. The server was booting from LUN0. The C-volume was very undersized at 6GB and the customer wanted to expand it to 50GB. All LUNS are presented to the server as separate SCSI disks, so we needed to move the C-volume to a new LUN created with the desired size.
We started the procedure by stopping all Exchange services. We then performed BareMetal backup of the server. After this was completed, we shutdown the server. The customer created a new LUN of 50GBs, we unmapped all LUNS and then remapped them replacing the old 6GB LUN with the new 50GB LUN in the first position. All other LUNS were mapped in the same order as they were before. To the server this now looked as if we had pulled a 6GB drive and replaced it with one that was bare and 50GB in size. We then booted the server from a Windows PE BMR CD with the appropriate Qlogic driver injected (when injecting this driver, the mass storage option should not be checked). We then restored the BareMetal backup to what was presented as disk 0 (our 50GB LUN). After the restore completed successfully, we ran the following commands from diskpart:
select disk 0
select volume 1
We did not use the size value as we intended to expand the volume to use the entire disk/partition.
We then rebooted the server to Windows. There were no errors and Exchange started up just as expected. Exchange was off line for just a few minutes. Worked like a charm.
With the customer, I used this to teach them the BareMetal restore process. We were able to perform tasks on production servers that they had been putting off for months because they thought it would be very difficult and time consuming. This is a customer that will be a very good reference account for us.