Mark of technology

Recently we had a support customer with a problem server - in fact, it was their only server (a small business customer) - and their server was in real trouble!

Originally reported to us as problems accessing certain services (database, email) on the server, with others (files) being painfully slow, we attended site and discovered a problem with the server RAID array - in fact, two of the drives in their RAID 5 array were failing - the only reason why such a problem had not yet knocked out the array completely was that for some reason, the controller was simply trying to carry on, repeatedly attempting to rebuild.  Worse still, the customer had no backup(!)

The server was painfully slow - rebooting took all night, and when we got the server back to our offices for more 'involved' work - primarily taking an image, it took 24-hours to complete, despite only being a small image of around 60GB.  Our imaging process also reported various errors, so we couldn't be sure of the integrity of the image itself.

We pushed on and tried to restore to a 'spare' server we had, which we'd loan to the customer if we could get it working.  It was not the same hardware, and we're talking Windows here - but we were using imaging software that could work around the problems of dissimilar hardware - but our plans were thwarted - we were fought every step of the way, with the server refusing to do anything other than give a BSOD on bootup - numerous approaches to solving this problem were tried, including repairs and re-installs of the OS, to no avail.

Whilst this battle was going on, I opened up a new front of attempting to use VMWare converter to make a virtual machine out of the image.  I'm not exaggerating when I say this; it worked first time.  And when I say it 'worked' - i mean it worked!  Their Exchange databases mounted without a hitch, same too for their main database, all we had to do was reset the IP address and it was done!  From that point, the solution was clear!

We gave up on the direct hardware restore, and instead installed the free ESXi 4 on our spare server, moved the image over, tested it was still working (it was), and then deployed it to the customers site, where it works perfectly, providing a much needed stop gap measure whilst their original server is evaluated for repair or replacement!  Happy customer, and a big problem solved for us!

Could we have done this with something other than VMWare ESXi?  Such as Microsoft Hyper-V or Citrix XenServer?  Well, possibly, but the first problem with Hyper-V or XenServer would have been its insistence on hardware assisted virtualisation.  We've got plenty of 'spare' servers around - but many of them are older systems with CPU's that don't support hardware assisted virtualisation, it would have narrowed our choices significantly, and lengthened the process of getting a spare system up-and-running.  By comparison, with ESXi, we simply had to concentrate on making sure the server had the basics covered - RAM, hard drive space etc - much easier!

Needless to say, we'll certainly be looking to use a similar approach the next time this happens - and VMWare will most likely be our platform of choice to send to the rescue!