Something I talked about yesterday at the Phoenix SBS 2008 build day is specing the servers appropriately. There was a thread on the SBS2k yahoogroups listserve that showcased the issue with a server that wasn't quite up to the task.
Fred Limmer wrote... "A while back I posted about some problems with two SBS 08 installs that
I performed simultaneously. Both installations were exhibiting the same
problems about a week after installation was finished.
Both boxes had identical specs (I have about a dozen servers that are
almost identical to these(slower procs and less RAM) that are running
SBS03 without a hitch):
Dual Core2Quad Xeons 2.0GHz
4x500GB HDD in RAID5 Hot swap bays
Intel V5000 Chipsets
Intel RAID controller
Services would refuse to stop, start, or would hang
Backups would not complete
Servers ran very sluggishly, particularly under use, and occasionally
Errors in logs from a variety of sources, but nothing that clearly
indicated the problem
VSS Writers would time out and not report status to vssadmin using
"vssadmin list writers"
After 3 months of working with PSS on these issues and reinstalling,
restoring, recreating RAID arrays, moving SQL, Exchange, WSUS Databases,
countless hours on phone with PSS, and eventually even swapping for
completely different hardware TWICE, we have finally found the source of
Having the RAID5 array on 4 hard drives was the source of the issue.
Apparently the controller was not able to keep up with the read/write
demands of SBS08 spanning across 4 drives.
On the backups, it was taking too long to create VSS snapshots of the
array due to having 4 spindles. From what the SBS Engineer (that I was
handed to after a couple of days working with level 1 PSS), explained to
me was that the SBS Exchange VSS writer, which, as you know is specific
to SBS, was timing out It takes 3-4 seconds per spindle to create a VSS
snap, and the Exchange writer waits 10 seconds after the initial call to
run its process. With 4 spindles, that's 12-15 seconds to finish the
volume snapshot. By that time, the Exchange writer has given up and
shut down. The writer doesn't finish prepping the stores and the backup
To make a long story short, adding two drives in a mirror on a second
RAID card and moving the Exchange stores to the new volume resolved the
The lesson here is, SBS has crazy high performance needs, and you
should definitely use a bad-a** RAID controller, or use a configuration
that spreads the load out more like 5 drives in a 2XRAID1/3XRAID5, or
something similar, but do not use 4 drives in a RAID5. It's nothing but