Hello, we are having problems with our load...
Server 2003 GER SP2
Citrix Presentation Server 4.5 R02
After rebooting our server the load often stays at 10000.
The only thing that works is to stop ima and sma service
type dsmaint recreatelhc and start ima and sma service again.
Unfortunately I could not find any solution unitl now and the problem is getting really worse.
When I check the citrix load evaluators the cpu shows 100%. Comparing this value to perfmon on the server it is not correct!
Thank you !
This is by design. Shortly after a reboot, the Citrix load evaluator will report 100% load (10000) to allow all the services and system components to reacha stable state. This prevents the "Black Hole Effect" where a freshly booted server suddenly starts taking every new connection and effective grinds to a halt trying to handle the requests. If you wait a few minutes, you should see the value start to drop. What Load Evaluator are you using?
Why is it called "Common Sense"? It doesn't seem all that common!
we'll unfortunately the load 1000 stays for more than one day...that can not be normal! We use CPU, Memory and Users as Loadevaluators...
We reboot our servers weekly and often more than ´50% stay at load 10000
The problem seems to be with the cpu values...citrix says 100% cpu load which is wrong ... somehow the value is not updated
OK... check some registry entries on your servers. I ran into this problem a while ago, and it had me stumped for the longest time.
Look under HKLM\SYSTEM\CurrentControlSet\Services
There should be four different sections you'll want to check out:
PerfDisk, PerfNet, PerfOS and PerfProc
Look in those, and under the Performance section within them. Look for any value called "Disable" or something similar. If there IS a disable value in there, delete it. If any one of those entries is disabled, it will completely screw up your Advanced Load Evaluator settings. It's a stupid windows dependancy.
thanks for your advice! Unfortunately I don't have an entry with Disabled in the regkeys...
any other ideas?
How long did you say it takes for the load evaluator to go back to normal after restarting the server? It shouldn't be longer than 2 or 3 minutes tops. Also, how many servers are there in this Citrix farm?
Do you have multi language support enabled (MUI)?
If so, have a look at this article:
Another possibility is that you have inadvertently added a scheduling metric to the LE assigned to the problem servers.
President (MCSE, CCNA, VCP, CCA)
VCIT Consulting - Citrix/Terminal Services Remote Desktop Solutions for SMB
VCIT website My Blog
Hi Alan, no I already checked that...we don't use MUI
are you running on VMware ESX?
Had this issue on a server after upgrading VMware tools.
Solved it by uninstalling VMware tools and doing a manual installation without the "Volume Shadow Copy Services Support" component. (The issue remains even if you modify VMware tools to uninstall the component)
Before finding this out I tried recreating Performance Counters and replacing the LMS20Rules.dll with one from a previous HRP without any success.
Hope this helps!
Hi, no they are physical servers...
are you running any antivirus on the servers?
If so, did you exclude the c:\Program Files\Citrix + all subfolders and files from being scanned?
My thoughts are that the antivirus is messing with your LHC....
Have you run DSCHECK /full servers /clean?
You have not installed MUI on your servers, but I thik it would be worth a try setting the translation registry seting.
HKEY_LOCAL_MACHINE\Software\Citrix\IMA\LMSName: EnableTranslationType: REG_DWORDData: 1
If nothing helps, try C4 :)
No, seriously, I would opt to creating a custom loadevaluator rule leaving out the CPU while continuing the search for answers.
Hi Miseh, thank you very much for your answer and recommendations!
I noticed on one server where the load absolutely didn't change from 10000 anymore that the windows performance counters where messed up. I could only see numbers and no naming anymore.
As citrix uses the windows performance counters for it's load evaluators I now tried resetting the values as described here ... http://support.microsoft.com/?scid=kb%3Ben-us%3B300956&x=13&y=10... on two of my servers.
Maybe there is a problem somewhere here...I will check the next two weeks and let you know if the problem is solved like this
Thanks again for your help!
I have had an issue with missing performance counters a few years ago.
To solve it I exported the correct registry entries from a healthy server (same model) and imported it on the faulty server, rebooted and everything was ok again.
In your case missing performance counters is not the problem since you get values when running perfmon.
And I don't think rebuilding them will help you since you are able to "resolve" the issue by recreating LHC....sorry.
I am starting to think that the problem is in your DataStore.... or that you are missing some important hotfix.
You should try bumping one of the servers up to HRP03 or even HRP04 just to see if the issue remains. (if you do, don't forget to read the prerequisites)
I have seen both the MUI corruption issue and local counter corruption KB300956 (as referenced by Oliver), basically you need to run the lodctr /R command
The first thing to check is, do the Perfmon counters work locally? Fire up perfmon and see if you can get any data out, the next thing is to create a custom evaluator and see which counter creates the load of 10000, e.g. add users, context switched, CPU, Mem etc.
I wanted to thank you for your advice. I had a VMware virtual machine with this problem.
As you said, it was fixed by removing VMware tools. (I did it from the add/remove programs).
I don't have enough words to thank you, because i was dealing with this problem for a long time.