I believe I have a server with faulty hardware. I think it's the motherboard, because I've tested the RAM, but anyway, I'll explain what happens.

My exchange server has 1 Private store (13GB), 1 public store (8GB) Raid5, exchange stores are on a separate partition which is 38GB in total size, about 10GB free space. It's a 32 bit OS (win 2003 server standard), with 4GB of ram. It has 2x E540 @ 2.50GHz processors

Every night between 1am and 5am, Exchange performs online maintenance on its exchange stores. The problem is that every night, while one of these maintenance procedures is executing, the server freezes/locks up. When I get to work and check on the server, the server doesn't respond and screen is just locked on a stuck screensaver.

I've only figured out why it is freezing because I logged on at 1am and watched taskmgr and saw that a particular process was chewing through a lot of ram, and then just suddenly I lose connection to the server. It turns out to be exchange.

So my question is: what does exchange do while in this maintenance period? I thought it was basically just a defrag. Which, if that is the case, I might have to move the exchange database to another computer and perform an offline defrag every month or something.

If you give this server a mammoth task to do.. involving a lot of memory.. it freezes. HP don't think anything is wrong with it after their remote testing and don't want anything to do with it. I'm raging.

Any thoughts or suggestions would be greatly appreciated.

  • How big is the Mailbox Store, how many Mailbox Stores, how much RAM, what processor and what disk config is in the server? – Ben Pilbrow Jan 21 '11 at 20:08
  • 1 Private store (13gb), 1 public store (8gb) Raid5, exchange stores are on a separate partition which is 38gb in total size, about 10gb free space. it's a 32 bit os (win 2003 server standard), with 4gb of ram. 2x E540 @ 2.50GHz processors – Mike Jan 21 '11 at 20:46

The Online Defrag process is both memory and disk-io intensive. It isn't particularly CPU intensive. In terms of stress, I'd hazard a guess that it's harder on the I/O subsystem than it is on RAM. That said, your own testing suggests big-RAM processes will cause the hard lock. Hard-locks in Windows are typically very low level faults of some kind, which does suggest hardware in some way. It may not be Memory exactly, it could also be the memory sockets, some heat fault in the chipset (very unlikely in an HP server, but it could happen), or even CPU.

These are hard to isolate, especially since it seems load related. Motherboard is not a bad diagnosis.

| improve this answer | |
  • Yeah, I've been having a hard time tying to figure out the problem, especially because it's a production server :( – Mike Jan 21 '11 at 20:56

I wouldn't be inclined to point the finger at Exchange - the server you describe is more than adequate for the job of running Exchange 2003. To give you a comparison, we have an Exchange 2003 server of similar spec, and the Mailbox Store is more than 10x larger and we have no issues during the online defrag process.

The maintenance period tasks are documented in this Microsoft KB (the defrag being the most resource intensive). Since you say any "heavy" work will bring the server to its knees and not only the Exchange maintenance period, I'd also point the finger at some faulty hardware.

I wouldn't recommend doing monthly (or any really) offline defrags of the Mailbox Store, it will only annoy people. The store must be dismounted for an offline defrag (so no email during that time) and it will take a long time.

| improve this answer | |
  • I know it annoys people, but the other option is to turn off and not bother with maintenance at all. I have had to turn off the maintenance plan that runs at 1am. I think new server hardware might be needed. – Mike Jan 22 '11 at 20:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.