Exchange 2010 sp1 mailbox, hub, cas roles and w3wp.exe / very slow

  • Thread starter 787or380
  • Start date Views 2,585
7

787or380

Hi,

Same issue as per http://www.experts-exchange.com/Software/Server_Software/Email_Servers/Exchange/Q_26508760.html

I am running Exchange2010 SP1 latest update, w2k8 R, a dedicated server with 4 quad core processors, 16GB or RAM and iscsi 15000rpm disks. 100 users, 40 active mailboxes. After a day the Exchange is slowing down until OWA (server is busy)and RPC connections fail more and more frequently, solution, rebooting the box.
Others solution found recycling the the different Exchange apps pools every 6 hours or killing the w3wp.exe processes.

I have installed 2 fixes for 2k8 R2 memory leak but no joy. I run a perfmon and the exchange troubleshooting tools: result all ok, no bottleneck, no high disk I/O, ecetera.

The Exchange transition is finish but the 2003 is still part of the org.

Any idea?
 
R

Ranger 989

787,

How does the following counter look in the Performance Monitor in the Exchange toolbox. RPC Latency Average (msec) in the MSExchange Store Interface. Let me knwo what the numbers read.
 
7

787or380

Hi,

No Antivirus on the box.

RPC latency around 7.100 stable on a 10mn period (9:30am, heavy traffic at the time).
 
7

787or380

We got some latency again, the recycling is not working.

But restarting the information store help, store.exe drop from 8GB to 2GB (this is normal), then everything works fast again, but I guess killing all the connection on the store can only help (except for the users).

I had setup a limit of 8GB for the store via AD object after I start encounter this issue but it did not help. I may lower it to 4GB making sure I still have some RAM left (16GB), because the average memory is 15.5GB on my Exchange box when the store.exe is 8GB.

On other thread I read that it could be a Domain Issue (GC), too many ldap connection on the GC, bad DNS settings? I will try rebooting my GC also DNS when the latency occur again. Then if occur again I will lower the IS usage to 4GB of RAM.
 
S

Steve Goodman

Lowering the cache for the Information Store will only increase your disk IO, it is not best practise to change this as far as I am aware. As an aside, is your page file size set to RAM+100MB?

Before you start rebooting domian controllers and such, have a look at the performance counters technet article . This should give you some idea of the values to check along with suitable thresholds. You'll even see counters for LDAP queries, which may help identify/rule out the domain controllers.

Edit: Also, have your run ExBPA on your server? Does it show any issues?

Steve

Steve Goodman
Check out my Blog for more Exchange info or find me on Twitter
 
7

787or380

Hi Steve,

Thanks for replying.

EXBPA was clear of errors since the install, Perfmon values using your recommended articles were also good. (Good to know this KB).

So far so good after rebooting my recently promoted DC/DNS/GC (installed before the Exchange build) use by the Exchange2010. I had some sysvol replication issue that resolved by themself after reboot, not sure if it was the cause of my trouble. Now the physical memory is almost around 15GB but no latency, hanging or slow EMC. I kept the store.exe cache size to 8GB.

I think that Exchange2010/2007 is now using a lot more IIS and if IIS cannot get the information immediatly from a DC/GC, the all things hang!
Per example the OWA logon page was always displayed instantly but when you tried login, it hanged, maybe trying to retrieve some info on the GC, same for Outlook.

I do not believe in a memory leak in IIS anymore, I just think the Exchange app pools use a lot more ressource and have more interraction with the GC.
As an other example when closing my EMC after a day of use, one a the w3wp.exe drop from 1.2GB to 600MB.
 

Similar threads


Top