Moving SBS2003 from a 5yr old server to a new one using StorageCraft

Body:

We normally do a swing migration for many of our clients when we upgrade from one server to another. This has the effect of minimising the downtime to the client and at the same time having a new server in place that’s faster and cleaner than the old one. StorageCraft with their ShadowProtect IT Edition product have now given us an alternate scenario that we can use to move a customer’s old server to new hardware, or even into a Virtualised system.

Now – I caution you – just like a swing migration, this is not a case of one shoe fits all. I’ve decided that I’ll only do a migration like this under the following circumstances;

  1. The existing server has performed flawlessly for 6 months or more
  2. The existing server has a non OEM license (it’s not legal to do it otherwise)
  3. The existing server can be removed from the customer site for a weekend

This particular customers server has been in place for 5 years plus. It’s an old HP (original HP, not post Compaq merger) Netserver E800 (Pentium III 800Mhz, 896MB RAM). The reason we’re upgrading the server is that the customer needs to run a SQL based practice management system that needs 3GB RAM and the current server can only take 2GB. The server itself has performed well for the customer and we were hoping to hold off till next year when SBS Cougar is released, but that is not to be now (the wait that is).

Ok – so we brought the customers server back to the office. Installed a USB 2.0 PCI card into it and connected up my 1TB external USB drive. Booted from the StorageCraft IT Edition CD and backed up the existing partitions (it was a software mirrored drive set). Back up took around 4 hours for the 36GB of data to come across. During the backup though, I had several blue screens of death (BSOD) around NMI/Memory Parity error which slowed me down a fair bit. I checked the spec for ShadowProtect and realised that I could do the image with only 512MB RAM, so by a process of elimination I found that one of the memory modules appeared to be faulty. Once removed the image went through without an issue.

With the image done, I moved my 1TB external USB drive over to the new server – a HP Proliant ML350 G5 with 2 x 146GB SAS drives, 4GB RAM and Dual Core 2.33Ghz CPU – a nice fast box. Booted the system, created a hardware RAID1 drive set (Mirrored), booted from the ShadowProtect CD and found that it recognised the HP RAID controller just fine. I found I needed to refresh the drive info though before my external 1TB USB drive came up. I proceeded to restore the images, and on the way through made the C: drive 30G in size (rather than 8GB that it was originally) and gave the rest of the 100GB+ drive space over to the E: drive. The entire image set restored to the system in just 13 minutes! Wow – that’s fast. I ran the Hardware Independent Restore option and it found most of the drivers itself – it needed drivers for the IPMI interface which I elected to skip for now.

Now – the tricky bits! Reboot the system, but boot it into Directory Services Recovery Mode (DSRM). Once in that mode, pick your internal NIC and assign it with the IP that your server originally had on the internal NIC. Also – you want to remove the original NIC’s drivers. If you bring up device manager you will not see the original NIC. You need to do a few things to make it visible – MS KB article 241257 shows you what to do. Once you can see the original NIC, delete it from the device manager and your good to go.

Reboot and cross your fingers. In my case we had a few problems. During the boot phase we got a SQLDUMPER error, indicating that the system had been tampered with – ok – I guess that it might think that given we had just moved the entire hardware platform underneath it. We got a few event log messages about this as well – see below.

Event Type:        Information

Event Source:        Application Popup

Event Category:    None

Event ID:        26

Date:            9/8/2007

Time:            8:21:03 AM

User:            N/A

Computer:        SERVERNAME

Description:        Application popup: Service Control Manager : At least one service or driver failed during system startup. Use Event Viewer to examine the event log for details.

 

 

Event Type:        Error

Event Source:        Service Control Manager

Event Category:    None

Event ID:        7022

Date:            9/8/2007

Time:            8:21:03 AM

User:            N/A

Computer:        SERVERNAME

Description:        The SQL Server VSS Writer service hung on starting.

 

I did a bit of googling and found that this was likely due to SQL 2005 being installed on this server. We had previously installed Dana Epp’s Firewall Dashboard software to monitor attacks on the ISA server and this had in turn installed an instance of SQL 2005 Express. Removing and installing it was not a big problem for me – but I wanted to understand the issue more. I decided that based on the error messages above that I would remove the SQL 2005 VSS Writer as I suspected that this was the issue more than anything else. Once removed and rebooted things were fine. This service gets installed to allow SQL 2005 databases to be backed up correctly by any VSS aware backup program.

 

Ok – given I removed the VSS writer, how did I fix it? Well – I needed to install SQL 2005 Express so that the new practice management software could be installed – it reinstall the VSS Writer which I verified by checking Add/Remove Programs and also "vssadmin list writers" at a command prompt

I also found that the exchange services didn’t start. I tracked this down to the fact that the original E: volume had taken up drive letter D: now – ok – easy to fix that, I reassigned the drive letters around and rebooted. We’re almost good to go at this point except for a funky Parallel Port error detailed below.

Event Type:        Error

Event Source:        Service Control Manager

Event Category:    None

Event ID:        7000

Date:            9/8/2007

Time:            10:41:04 AM

User:            N/A

Computer:        DCBSVR1

Description:        The Parallel port driver service failed to start due to the following error:

The service cannot be started, either because it is disabled or because it has no enabled devices associated with it.

 

Hmm – strange error. I decided to check the BIOS in the new HP server and found that the Parallel port is disabled by default. Nice one huh. I don’t need the parallel port per se, but I want to make the error go away, so I enabled it and rebooted the system – it discovered more hardware and requested a further reboot. Once up the event log is clean and we’re good to go.

 

All up a fairly painless migration. It certainly took less time than either a new build or a swing migration, and from the customers perspective it had zero effective downtime. Risk to the customer is also zero as the rollback procedure is to install the old server!

 

 

Category: Troubleshooting
Published: 8/09/2007 9:56 AM

Leave a Reply

Your email address will not be published. Required fields are marked *