I had one of those cases of bad timing recently when two things went wrong with a system at the same time. The system in question is a customer’s Windows Small Business Server 2003 server running on low end Dell server hardware – PowerEdge SC420.
The first problem was relatively simple and I felt pretty comfortable about it. The server has a Travan tape drive in it with 20/40 capacity. One day it just stops accepting tapes. Being familiar with Dell Technical Support and their troubleshooting scripts I:
- ran a cleaning tape through it five times (yes, five)
- tried brand new tapes
- read the existing tapes on another server
- updated the firmware to the latest on the Dell website
- updated the device driver as well
- cold booted the server a few times
Having done all of that I felt satisfied that it was a hardware fault and rang DTS. I stated the issue and my attempts to resolve it as above, and the operator immediately agreed that it was a hardware fault and would arrange a replacement and a technician to install it (the customer site is an hour away so may as well use the warranty labour services they paid for). Because the customer office was closed for the night I actually had to tell Dell to wait until morning, even though the warranty was 4 hour response. No one wanted to stay back and be there to let the guy in.
At this point I have to say I’ve lately been wondering what the problem is with DTS that other people are complaining about? I’ve used them a few times lately and its been smooth sailing every time. Maybe I’ve just been lucky and my next call to them will be a nightmare. Maybe I shouldn’t have tempted fate. Oh well.
So by mid the next morning the new tape drive is in and is accepting tapes, and I RDP in and kick off an ad-hoc backup. A few hours later the SBS backup report comes in to say it has failed. I was a bit worried at this point – have we got another faulty tape drive, is the problem more obscure than first thought, will I need to get Dell out again???
The backup would get most of the way through and then report in the log:
The requested media failed to mount. The operation was aborted.
The operation was ended.
I thought I should look further into possible software problems, and investigations took me to this article. I installed the hotfix, but still the backups failed. Taking another look at the error message and found another article in my searching. Now this article was interesting because I had already run a full backup of everything on the system to a backup file on a share on another computer, and the total size of that backup was 28Gb. The Travan tape drive has a 20/40 capacity (20 uncompressed, 40 compressed). The systems I’m used to working with don’t hold a lot of easily compressed data, so I usually assume that in this case 20Gb is all I will be able to back up. The 28Gb backup file is for everything on the system, and when you take away the data that isn’t backed up in the nighly tape backup (eg RIS, application install source, the VSS volume) it comes in under 20Gb anyway.
I found however that despite it being less than 20Gb of data on disk, it was too much for the tape to hold. I set about pruning more data out of the nightly backup definition, and found these handy exclusions.
SBS2003 ships with W2KSP4, WXPSP1, and if you’ve installed SBS2003 SP1 per the instructions will also have WXPSP2 source files in your ClientApps share. There is no real need to back these up to tape on a nightly basis. You could include them in a weekly or monthly set but really its not hard to download them again if they go missing.
If you’ve got other installation sources like Office 2003 you can exclude these as well, as they are easily recoverable from the original CD-ROMs.
The NTBackup catalogs stored in C:Documents and SettingsAll UsersApplication DataMicrosoftWindows NTNTBackupcatalogs51 start to add up over time. You could exclude the entire folder from your backups but I like to be able to pull the latest catalogs off tape due to past experience. What I do now is move the catalogs older than say a fortnight off to another folder and then exclude that other folder from the backups. Two years of nightly backups of this server was about 500mb of catalogs, which is a lot when you’re trying to trim the backup under 20Gb.
It goes without saying you can exclude temp folders, the pagefile, and probably even the local user profiles for your server admin accounts.
The article linked earlier also suggests that for Exchange you backup the Information Store but not the database files in the file system. If you’re pushed for space you can do this, but I would recommend backing up the database files weekly as well.
Finally the patch and service pack uninstall files in C:Windows can be excluded as well. If you’re running an SBS2003 box thats a few years old, and you’ve been diligently patching and installing the latest service packs, then this can add up to several hundred Mb of data (1.3Gb on this server).
I’ve discussed a few of these ideas with colleagues and there is mixed opinion on the wisdom of excluding certain things like uninstall files and backup catalogs. I’m of the opinion that if you’re recovering your system from tape onto bare disk after a disaster then you’re probably not going to be too fussed about being able to uninstall a patch from 18 months ago. Of course you can please both camps if you run a mixed backup regime on the server. For this customer we run the nightly tape backup with the types of exclusions mentioned above, and then every weekend run another backup to file across the network to a share on a pc, this time of the entire system. The stuff that is excluded nightly is not changed very often, so is quite safe to backup only weekly.
So there you have it, a vague SBS2003 backup failure message and some fairly basic steps to mitigate the cause of that failure. If you’re using a more advanced backup suite for your servers then you’ll probably find they provide much clearer warnings and errors and don’t lead you down the garden path quite as much as NTBackup can.