ExchangeMTA
asked on
Outlook is requesting data from the Exchange Server (I hate this message)
It's not the hardware. Got 100Mbit switches, my workstaion is P4 2.2Ghz.
I recently bought a new Compaq server. Very Nice, Dual 2.20Ghz Xeons, 2Gb Ram, 15,000RPM Drives.
Exchange Pub/Private stores are on Dedicated Raid 5 Array. Win2k OS is on a pair of 18Gb 15K software mirror drives on seperate SCSI Chan. Installed Win2k Server, SP3, Exchange 5.5, sp4
Added to domain, did the Move Mailbox method. Got all public folders, settings etc on new box.
Deleted old Exhcange 5.5 Server. All was fine.
After upgrading all servers to Active Directory (the Exchange server is an additional Domain Controller not member)
the response is Slow. Especially when I click on an email with an Attachment in it.
I bounce around in Outlook XP trying to get it to stall. (Outlook is requesting data from the Exchange Server)
Only seems to happen when I click an email that has an attachment.
Nslookup, and ping come back with icmp response instantly, not a name lookup thing.
(Outlook is requesting data from the Exchange Server)
Tried running perfwiz -v and tweeked it a bit.
This is a SINGLE server in a company, doesn't connect to any other Exhcange Server.
I have a FreeBSD box (postfix) that the Exchange server sends all Internet mail to/from.
I have only about 3Gb of Stores on Server About 75 Users.
Event Log:
Event Type: Error
Event Source: Perflib
Event Category: None
Event ID: 1010
Date: 5/1/2003
Time: 1:31:04 PM
User: N/A
Computer: XXXXXXXXXX
Description:
The Collect Procedure for the "MSExchangeMTA" service in DLL "C:\EXCHSRVR\bin\mtaperf.d
Data:
0000: 3d 01 00 00 =...
Q: What can I do to fix the above error. Tried Regsvr32 /u etc... no deal.
Q: Does this just do logging, or does it raise the threads for MTA?
Q: Any recommended settings for pefwiz -v?
Q: Event log shows the following:
Event Type: Error
Event Source: MSExchangeIS
Event Category: General
Event ID: 1191
Date: 5/1/2003
Time: 12:35:19 PM
User: N/A
Computer: xxxxxxxxxxxx
Description:
The total number of threads configured for the message store (46) appears to be too small for the current store configuration.
Using 0x3c threads instead.
(Which 0x3c in decimal (base 10) = 60. ) So What is this enough?
I recently bought a new Compaq server. Very Nice, Dual 2.20Ghz Xeons, 2Gb Ram, 15,000RPM Drives.
Exchange Pub/Private stores are on Dedicated Raid 5 Array. Win2k OS is on a pair of 18Gb 15K software mirror drives on seperate SCSI Chan. Installed Win2k Server, SP3, Exchange 5.5, sp4
Added to domain, did the Move Mailbox method. Got all public folders, settings etc on new box.
Deleted old Exhcange 5.5 Server. All was fine.
After upgrading all servers to Active Directory (the Exchange server is an additional Domain Controller not member)
the response is Slow. Especially when I click on an email with an Attachment in it.
I bounce around in Outlook XP trying to get it to stall. (Outlook is requesting data from the Exchange Server)
Only seems to happen when I click an email that has an attachment.
Nslookup, and ping come back with icmp response instantly, not a name lookup thing.
(Outlook is requesting data from the Exchange Server)
Tried running perfwiz -v and tweeked it a bit.
This is a SINGLE server in a company, doesn't connect to any other Exhcange Server.
I have a FreeBSD box (postfix) that the Exchange server sends all Internet mail to/from.
I have only about 3Gb of Stores on Server About 75 Users.
Event Log:
Event Type: Error
Event Source: Perflib
Event Category: None
Event ID: 1010
Date: 5/1/2003
Time: 1:31:04 PM
User: N/A
Computer: XXXXXXXXXX
Description:
The Collect Procedure for the "MSExchangeMTA" service in DLL "C:\EXCHSRVR\bin\mtaperf.d
Data:
0000: 3d 01 00 00 =...
Q: What can I do to fix the above error. Tried Regsvr32 /u etc... no deal.
Q: Does this just do logging, or does it raise the threads for MTA?
Q: Any recommended settings for pefwiz -v?
Q: Event log shows the following:
Event Type: Error
Event Source: MSExchangeIS
Event Category: General
Event ID: 1191
Date: 5/1/2003
Time: 12:35:19 PM
User: N/A
Computer: xxxxxxxxxxxx
Description:
The total number of threads configured for the message store (46) appears to be too small for the current store configuration.
Using 0x3c threads instead.
(Which 0x3c in decimal (base 10) = 60. ) So What is this enough?
We have the same issue....
Checkout this link http://www.tek-tips.com/gviewthread.cfm/lev2/3/lev3/15/pid/10/qid/689226
it may explain and help.....the main fix I tried is at
http://support.microsoft.com/default.aspx?scid=kb;[LN];Q264731
I have tried the fix for us (17gb store & 350 users) but so far it hasnt helped, although it says it takes an hour for each gb of store
If nothing eslse, it is consoling that it isnt a unique issue.....
Checkout this link http://www.tek-tips.com/gviewthread.cfm/lev2/3/lev3/15/pid/10/qid/689226
it may explain and help.....the main fix I tried is at
http://support.microsoft.com/default.aspx?scid=kb;[LN];Q264731
I have tried the fix for us (17gb store & 350 users) but so far it hasnt helped, although it says it takes an hour for each gb of store
If nothing eslse, it is consoling that it isnt a unique issue.....
The fix we tried didnt help. And now it is even getting worse....almost to the point of making outlook unusable. It is not a network issue as I am running a fluke & HP Openview and the network never gets above 10% utilization and then only on spikes. The NIC in the clustered exchange 5.5 server on adv. svr win 2k never passes 4% util. This is about to move us to sendmail as this seems to be a wide spread issue that is being ignored by MS.......
Ok, I fixed this a long time ago.
Couple of things to try:
I have a batch file I run called maint.bat, My stores are on the f: drive of my server (Raid)
I do the dir before and after to see the space difference, ususally 200+ Mb is recovered.
(All Exchange Services must be stopped before doing this)
f:
dir
eseutil /d /ds
eseutil /d /ispub
eseutil /d /ispriv
dir
Step 2: (All Exchange Services must be stopped before doing this)
=====
The other thing that did it was, run perfwiz -v (-v = verbose) tweek the threads, for processes, for me this is my ONLY server, it's a Dual Xeon with 2Gb Ram, stores are on their *own* Raid 5 array. Raise these vaules up a bit, experiment. Read up on what the values do and how they apply to your envionment. Example noone uses X.400 stuff, anymore, so set this at Zero. Mircosoft I think sets this default to 10 threads, Most of you dont need it.
Did it work. No Yes. Post back here.
Couple of things to try:
I have a batch file I run called maint.bat, My stores are on the f: drive of my server (Raid)
I do the dir before and after to see the space difference, ususally 200+ Mb is recovered.
(All Exchange Services must be stopped before doing this)
f:
dir
eseutil /d /ds
eseutil /d /ispub
eseutil /d /ispriv
dir
Step 2: (All Exchange Services must be stopped before doing this)
=====
The other thing that did it was, run perfwiz -v (-v = verbose) tweek the threads, for processes, for me this is my ONLY server, it's a Dual Xeon with 2Gb Ram, stores are on their *own* Raid 5 array. Raise these vaules up a bit, experiment. Read up on what the values do and how they apply to your envionment. Example noone uses X.400 stuff, anymore, so set this at Zero. Mircosoft I think sets this default to 10 threads, Most of you dont need it.
Did it work. No Yes. Post back here.
(Continued)
I can do 3Gb in 1700+ seconds. About 28 Minutes. eseutil /d /ispriv
What do you have for Hardware. (All my drives at U320 SCSI 15K Hot Plug)
O/S is on seperate SCSI Chan, software mirror cage.
I can do 3Gb in 1700+ seconds. About 28 Minutes. eseutil /d /ispriv
What do you have for Hardware. (All my drives at U320 SCSI 15K Hot Plug)
O/S is on seperate SCSI Chan, software mirror cage.
We are running on a Compaq CL380 cluster server. Basically dual dl380 with (2) 1.0ghz Xeons & 1 gb ram. It takes us hours to defrag our priv store (16 gb) so we are very reticent to do so as we operate 24 hours a day....Last time we tried it ran for 2 hours and said it would take an additional 22 hours to do so we had to stop it.....
I think we may have found an (hopefully the) issue. I have been monitoring disk access performance on the shared scsi array and it seems that every time the disk access time is high, we get req. data popup. I feel this issue may be performace related to disk access. I saw another post somewhere else that eluded to this as well. If I can validate this, then I would feel much better. We are about to migrate to 2 new clustered server using fiber channel scsi via Emulex LP scsi adapters to a coupe of brocade silkworms with a EMC clarion 400 disk array for shared storage so hopefully our storage access speed will increase dramatically. I need to do more research with regards to what perf. counters are important to exchange....
I'll post more as we progress thorugh this and what I find as far as disk access performance..and thanks for your info....
I think we may have found an (hopefully the) issue. I have been monitoring disk access performance on the shared scsi array and it seems that every time the disk access time is high, we get req. data popup. I feel this issue may be performace related to disk access. I saw another post somewhere else that eluded to this as well. If I can validate this, then I would feel much better. We are about to migrate to 2 new clustered server using fiber channel scsi via Emulex LP scsi adapters to a coupe of brocade silkworms with a EMC clarion 400 disk array for shared storage so hopefully our storage access speed will increase dramatically. I need to do more research with regards to what perf. counters are important to exchange....
I'll post more as we progress thorugh this and what I find as far as disk access performance..and thanks for your info....
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
We needed the san for File & print and for sql storage as well. The brocade allows us to segment the drive array so only the appropriate scsi cards see the proper data. We bought with the intention of having all 3 critical systems (email, f&p & sql) on a heavy duty, highly availble system We also plan on doing regular rolling snapshots of our info store so if we have a failure our restore time for 20gb priv edb would only be about 10 minutes....
The servers we wil use to run exch 2000 on are compaq 8500s with (6) 550mhz xeons each & 4 gb or ram running clustered...
We had them in house already due to an aquisition.....Our store.exe is only using 400 mg of ram so I dont think ram will do much.....
We will try the /ds & /ispub over the weekend.....
Thanks, I'll post the results....
The servers we wil use to run exch 2000 on are compaq 8500s with (6) 550mhz xeons each & 4 gb or ram running clustered...
We had them in house already due to an aquisition.....Our store.exe is only using 400 mg of ram so I dont think ram will do much.....
We will try the /ds & /ispub over the weekend.....
Thanks, I'll post the results....
2Gb of ram (is to) 884Mb used by store.exe
AS
1Gb of ram (is to) 400Mb used by store.exe
That is the whole point. Store.exe will attempt to use a Percentage of total ram.
Having store.exe use much more ram *will* increase your performance. Ram is faster than any disk.
So if you follow this fairly linear equation...
4Gb of ram --> store.exe will use about 1.6Gb dramatically increasing speed to the exchange server and possibly getting rid of your
Outlook is requesting data problem..
(Add some more ram, its very cheap now)
http://www.ec.kingston.com/ecom/configurator/modelsinfo.asp?SysID=+9357+&distributor=0&submit1=Search
Look at the above link...
AS
1Gb of ram (is to) 400Mb used by store.exe
That is the whole point. Store.exe will attempt to use a Percentage of total ram.
Having store.exe use much more ram *will* increase your performance. Ram is faster than any disk.
So if you follow this fairly linear equation...
4Gb of ram --> store.exe will use about 1.6Gb dramatically increasing speed to the exchange server and possibly getting rid of your
Outlook is requesting data problem..
(Add some more ram, its very cheap now)
http://www.ec.kingston.com/ecom/configurator/modelsinfo.asp?SysID=+9357+&distributor=0&submit1=Search
Look at the above link...
Ok..now i get it....And after more research, it seems that the disk usage is high for the exact same reason, page faults per sec are high (16 to 50 with spikes to 100 on a regular basis)
I will add more memory, we probably have some in our stock of stuff....
I will let you know the results.....
Thanks again....
I will add more memory, we probably have some in our stock of stuff....
I will let you know the results.....
Thanks again....
Well...I have loaded my server with 4gb of ram, added the /3GB switch to my boot.ini per Microsoft. It seems that store.exe doesnt grab the memory unitl it thinks it needs it. Currently offload it is only using approx 700 mg of ram.....At least pages/sec are now under control....
We will see the results Monday morning when the system gets loaded......I will report back then....
We will see the results Monday morning when the system gets loaded......I will report back then....
Cool I think the Ram will help a lot. I actually came in over weekend and did an offline defrag of my stores...
Did you try your directory and the public stores?
Also, as users start to use exchange the store.exe will start grabbing more ram, I'd take a look at it around 2:00pm...
How's the lag? (if any)
Regards...
Did you try your directory and the public stores?
Also, as users start to use exchange the store.exe will start grabbing more ram, I'd take a look at it around 2:00pm...
How's the lag? (if any)
Regards...
I didnt get the chance to defrag anything. We are going to do that Wednesday night during our monthly sched. maint. window.
I am tracking the ram and it seems to be inching up as users show up to work. There doesnt seem to be a lag right now although I am still getting the occasional req. data. I am however now running outlook 2003 as our first test install so it may be more sensitive to it....
Thanks, and I will keep posting as the day progresses.....but it looks good so far....
I am tracking the ram and it seems to be inching up as users show up to work. There doesnt seem to be a lag right now although I am still getting the occasional req. data. I am however now running outlook 2003 as our first test install so it may be more sensitive to it....
Thanks, and I will keep posting as the day progresses.....but it looks good so far....
Sorry it took so long to post a follow up....we got swamped with new customers (a nice problem to have)
Here are the results of what we tried:
Added 3 gb of ram, tweaked memory used by exchange 5.5 (by running optimizer in advanced mode to maually set meory for exchnage to use 70% of physical ram, defragged public store (couldnt do the private due to downtime contraints).
Results:
Req. data went away for about 3 days, then returned. Several users with larger mailboxes (800mg) were even unable to function in Outlook due to req. data..
We have now migrated to an exchange 2000 environment on a pair of Compaq 8500 running 6 1ghz processors & 4gb of ram. Everyone is much happier and no one has recieved requesting data, except me....I am running outlook 2003 and I believe it is more informative (sensitive) of req. data than earlier versions.
We found a piece of software that really allows us to see exactly what is going on with our exchange environment. It is called Spotlight on Exchange by quest software ( www.quest.com ). The best thing is they have a freeware version that does a ton of stuff such as monitoring disk queue reads, mta connections, memory usage and all kinds of performance numbers.
Thanks for all the help....
Chris
Here are the results of what we tried:
Added 3 gb of ram, tweaked memory used by exchange 5.5 (by running optimizer in advanced mode to maually set meory for exchnage to use 70% of physical ram, defragged public store (couldnt do the private due to downtime contraints).
Results:
Req. data went away for about 3 days, then returned. Several users with larger mailboxes (800mg) were even unable to function in Outlook due to req. data..
We have now migrated to an exchange 2000 environment on a pair of Compaq 8500 running 6 1ghz processors & 4gb of ram. Everyone is much happier and no one has recieved requesting data, except me....I am running outlook 2003 and I believe it is more informative (sensitive) of req. data than earlier versions.
We found a piece of software that really allows us to see exactly what is going on with our exchange environment. It is called Spotlight on Exchange by quest software ( www.quest.com ). The best thing is they have a freeware version that does a ton of stuff such as monitoring disk queue reads, mta connections, memory usage and all kinds of performance numbers.
Thanks for all the help....
Chris
Hmm, Interesting. I too have upgraded my server to Exchange 2000. I did an in place upgrade. I forgot to setup a Connection Agreement. (Ouch!) It took a call to PSS to get it fixed up. Glad to hear all is running well.
Thanks for the tip on the quest software. I'll have to take a look. I guess in Exchange 2000 there is no "perfwiz -v" at all?
I did get the Outlook is requesting data, Message once on my server, but it went away after store.exe was using more than 1/2 Gig of Ram... I use the Symantec AntiVirus for Exchange 2000, has a big improvement over old version. Wow.
Thanks for the tip on the quest software. I'll have to take a look. I guess in Exchange 2000 there is no "perfwiz -v" at all?
I did get the Outlook is requesting data, Message once on my server, but it went away after store.exe was using more than 1/2 Gig of Ram... I use the Symantec AntiVirus for Exchange 2000, has a big improvement over old version. Wow.
I will add 1 more comment since this thread seems to be retiring. We did get our email server up to over 4gb (6 gb & 6 processors) and still had requesting data. The issue appears less related to memory but associated with mailbox size. Once we got all mailboxes below 300 meg, the issue appeared to have gone away. The only time it ever came back was when individual mailboxes got larger than 300 meg. Only the users with very large mailboxes get requesting data now and only when their mailboxes get larger than 300 meg....the minute we gte them to archive their mailboxes, the requesting data stops.
ASKER