NIC Queue length 4 billion?

[lwolfson]

I realize this my be a NT4 Server question, but maybe has something to do with E5.5:

Server: NT4 SP6a/Exchange 5.5 SP4
Network utilization is low: This server is in a collocation with 10M pipe to Internet. I have rarely seem spikes above 2M at any time.

When I check the NIC Output Queue Length in Perfmon is shows a count of over 4 Billion!

Rebooting does not clear it and there is no mail stuck in the queues.

Any ideas out there? My first thought was to shut down the store and IMC to see if it goes away to see if Echange is trying to push out something, but that's not it.

Anyone else had this happen to them?

 

[Taffyboyo]

Are you sure this is a problem?
Are your users experiencing problems?
The reason I ask is I have run the same performance monitor on our server and it to runs at a constant.

 

[lwolfson]

Yes
Possibly
What is the Queue length you see?

The last time I saw this was with a big production web site collocated at 3 locations each on 10M pipes: the web site launch required a 7M download and the lines were saturated in the UK and we saw a queue of 5! Just five, not five billion. We upped the band width to t3 in NY and SF in time to avoid the logjam. Also used redirection to take load off UK server and balance to SF.

Still, 4+ billion is NOT normal. Line is not saturated and all traffic is fine. Users experience and occasional hang due to RPC timeout when GAB is queried on a send from Exchange.

 

[SirTravis]

I have this same problem and cannot find an adequate explanation for it yet. Have you been able to solve it?

It can't actually be THAT high because the server is still working, albeit a bit more slowly than I would like. The 4.2 billion value must be a misconfiguration or something. I would just like to know what it really is.

BTW> I'm running windows2000sp2 on the server.

SirTravis the perplexed

 

[lwolfson]

Thanks for the info. MS recommended I rebuild the OS to 2K and install 5.5 back in and restore pub priv and dir.

I suspect two possibilities: a large number of folks with 1000M+ mailboxes syncing all the time and huge size of dir (due to 30000+ public folder, each of which get an SMTP address automatically).

I will rebuild the server on July 4 and post the results. Certainly hope it clears the queue.

 

[lardum]

This is normal behavior on a SMP machine. I remember when i investigated this behavior long time ago and found out that this happens when you have 2 or more CPU's in the machine. I don't remember what the explanation was. However the machine shouldn't be affected by this. You should still have good performance.

 

[msw8738]

Check out this microsoft support page. If you have a symmetric multiprocessor server then this counter, and others of the same category, are inaccurate.

SNMP Counter Values Are Unreliable in SMP Computers

http://support.microsoft.com/defaul...port/kb/articles/q146/0/04.asp&NoWebContent=1

 

[SirTravis]

Thanks msw, that link explains it. I knew the value had to be wrong but I just couldn't figure out why. It's annoying that the counter is useless but at least now I know it is.

 

[lwolfson]

Thank you for finding this proof.

I guess I will have to look elsewhere for the cause of directory hanging up, but the more I hear, the more I find that common among those who run Exchange remotely.

Thanks again, all.