Memory Leak in Exchange 5.5 SP4

[Electro]

Hi All,

Recently we have noticed that our exchange server is having a lot of problems with Memory Leaks. We shutdown the Information store and related services and the memory becomes free again but then slowly leaks once the services have been restarted.

We are running NT 4, SP6A, MS Exchange 5.5 with SP4.

Anyone else had this problem or can provide a fix would be greatly appreciated.

Thanks,

Darryl Brambilla
EDS Canada
Darryl.Brambilla@eds.com

 

[absynther]

I am having the exact same problem. At first I attributed the leak to not being able to get a service pack 4 install on exchange 5.5 to take. After re-installing OWA to get Exchange service packed to 4 on an NT4 sp6a with the OWA script(ver3) and ida/idq.dll hotfix installed it still leaks memory. All threads regarding memory leaks point me to exchange sp4 with nt4 sp6a. I have followed all correct heuristics in service pack application order, yes I applied sp6a at all the multiple times it is required. I am fearing my only recourse is a full removal/re-install of exchange. SOMEONE PLEASE TELL ME I'M WRONG.

 

[DTwo]

Same here,
Followed all sensible routes, I'm now considering wiping and starting from scratch with sp3, As i have spent more time/downtime (due to crashes/out of memory) trying to solve this rather than losing the lot.

It is annoying to say the least

 

[tyrekicker]

Have simialr problem but just get crashes on regular intavals when store.exe uses 100%CPU. I have done one full reinstall and made mistake of going back to Exchange SP4 and still have problem. Been through over 200 technet articles regarding store.exe and processor usage. Eseutil & Isinteg are now old friends! Is my problem related?

 

[DTwo]

Tyrekicker,

The problem i'm experiencing isn't related to the store itself or it's cpu usage.

Actually, whatever it is that is growing to use all the ram, is not identified in the taskmanager, it's system related being provoked by exchange (if that makes any sense).

I do know that it is related to tcp/ip activity in exchange....almost like a slow DOS attack...from the inside

Anyway,
I'm starting to get a lil fed up with having to restart all the exchange services once a day to clear the memory before it takes the server with it......

now if i could only find the time to redo it all.....

 

[Maruis]

Hi Guys
I have the same server setup.
My server uses all the memory I gave it. THAT is normal, remember Exchange has a database and it will use all the memory it can.
What happens is that it tries to load all your mailboxes into memory for faster access. It looks like a memory leak but it aint.
Do this test to see it for yourself.
Reboot your system then open the largest mailbox you have or even a couple of them. You will notice that your memomory usage will increase dramaticly.

To solve your problem you need to add memory to your system the more the better.

Hope this answers your questions.

 

[BGSUgirl]

I don't know if this is related, but are any of you running NNTP (Network News Transport Protocol)? There is a new Microsoft Security Bulletin that talks about how that service in Windows NT 4.0 and Windows 2000 contains a memory leak in a routine that processes news postings. If someone sends a large number of posts, the server memory could be depleted to the point at which normal service would be disrupted. Here is the url in case any of you possibly find this useful:

http://www.microsoft.com/technet/security/bulletin/MS01-043.asp.

This may not be related to your problem, but I thought I'd at least mention this bulletin.

-Lauren

 

[Maruis]

You are right on NNTP, however I do not run NNTP at all.
Thx for the reminder

 

[DTwo]

Maruis,

Thank you for your response, I do understand that exchange is ment to use availiable memory.

The server in question has run happily for 1yr+ on the current amount of installed ram and the store it's handling at the moment is more than 50% smaller than it has been in the past.

What I am referring to is a leak that bring the system to it's knees,
eg
Event ID 2019
The server was unable to allocate from the system nonpaged pool because the pool was empty.

I'm not talking about something that "looks" bad memory wise in the task manager or performance monitor.

We're talking knock the system down, blue screen of death, "I ain't working no more" crashes, after half a dozen ID 2019's.

It takes 12-16hrs and doesn't seem related to the actual volume of mail traffic. Restarting exchange related services resets the "clock" so to speak

I've checked everything I can find on 2019's and none of them are actually related to my situation/setup or exchange.

This problem is exchange though, the system is perfectly happy without it......but that kinda defeats the purpose of the server though

Throwing more ram at this server is not going to solve my problem

I don't run NNTP either i'm afraid.

 

[Maruis]

OK
This changes the whole problem, you have a winsock problem and not a exchange problem.

Use this URL will give the KB on this specific problem:

http://support.microsoft.com/suppor...D 2019&rnk=5&src=DHCS_MSPSS_gn_SRCH&SPR=NTS40

Hope this help you guys

 

[DTwo]

Thank you for your further response Maruis,

The article you mention did come up quite early in my investigation into the problem.

Unfortunately the symptoms MS suggest,

eg

"You can observe this problem with Performance Monitor. Log the Non-Paged Pool Bytes item under Memory Object, and watch for the value to grow to 20-40 MB."

Were not apparent on the system.

I'll check this again over an extended amount of time (obviously less than 12hrs just in case.....it does seem to be something else though.

I do appreciate your suggestions

 

[Maruis]

Sorry to hear that...
There is more you can try out if you want..
Reinstall servicepack 4 for exchange, reboot and reinstall servicepack 6a for NT, reboot.
Also make sure that the conections you don't use is disabled.
I really do not know whether this will solve the problem, you got nothing to loose so why not try it.
In the past I had instances were reinstalling the servicepacks solved some weird problems.

Let me know what happend

 

[DTwo]

Thanks Marius,

Reinstalling the service packs in that way does seem to have done the trick.

The server had been running several days without dropping dead or needing the exchange related services to be restarted.

These service packs had actually been restored during this little escapade before, I'm not sure why this has worked......just love these "process" related fixes

Thanks again

 

[Jackylee]

Hi, maybe you can try to improve the performance of your system by improve virtual memory.

Hope this can help you!

 

[Maruis]

Glad to be able to help you out in your hour of need.....