Well, what I’ve seen before is some muppet has configured ICA connections to connect directly to the server rather than load balance – especially if the server was one of the first in the farm. So because you’ve a different number of load evaluators you can end up with more users on that servers because x users *always* log on to that server – you might what to check the name of the icaconnections - do they all have the same name? No blanks?
You may want to read http://support.citrix.com/article/CTX111035 ... or you may not as it may be you...
I think in the first instance the command queryds /table:LMS_ServerLoadTable would be your chap – that’ll give you a break down of which evaluators are weighted where and how – that’ll give you more information than simply doing a qfarm /load. Queryds is on your Citrix CD, I don’t think its installed by default.
You’ve got 4 evaluators
Application User Load, - Thats a straight count of users for an application - whats the application? Is it a ‘core’ application i.e. is your published app the only app on the server? Can the users launch other apps and still have a connection without this app running?
Load Throttling, - The default is ‘high’ – but I’d have said the impact of this evaluator would only kick in during peak logon/logoffs – less likely to impact over time... but .. meh ... open mind an’ all that.. whats yours set at?
Memory Usage, - This is a character building evaluator – unless you’ve got apps that use a regular memory footprint it can cause all sorts of “funny” skews on user count – especially if you’ve not got optimisation on, or you’ve not throttling user memory consumption. Flip side is – it stops you running low on memory available on the server
Server User Load – straight user count
So – it may be application load and server user load have cause to be different over time – say you base your app user load on Outlook, but the user launches outlook, then launches word, then closes outlook. Still got a user on, and your memory and app user load evals are going to be lower – so next user that comes on will go to this server rather than another (where users are running outlook) – over time that’d cause some differences
It could be that there is a different process running on that server – CTXPRESS1P – or that the memory optimisation is only configured to run on that server – or, its been running on that the longest so the overall memory optimisation.
I’d be tempted to suggest to cut out the application user load in the first instance & load throttling if possible, and trouble shoot with queryds from there – but regardless run queryds /table:LMS_ServerLoadTable should show which rules contribute more
If you’re still having problems look at the running the scripts suggested in the KB article
a.
From: thin-bounce@freelists.org [mailto:thin-bounce@freelists.org] On Behalf Of TBarnhart@rcrh.org
Sent: 23 February 2009 23:47
To: thin@freelists.org
Subject: [THIN] Re: Citrix Load Evaluator / Total Sessions per Server & Application
Andrew,
Yes.
Well, actually it depends on how the "alphabetize" process was programmed by Citrix...
(You know the whole 'left-to-right' issue w/ 1, 10, & 01 in numbering.)
In the whole 4.5 farm: No, about 1/3 of the way down the list of 46 total.
In this set of servers w/ published-app: Yes, #3 in the list of 20.
RUS-CTXPRESS10P
RUS-CTXPRESS10S
RUS-CTXPRESS1P <<<<<<
I seem to remember that our Support asked that question once, but it never went any farther…
Any ideas on where to track down this train of thought?
Thanks,
Troy
-----Original Message-----
From: thin-bounce@freelists.org [mailto:thin-bounce@freelists.org] On Behalf Of Andrew Wood
Sent: Monday, February 23, 2009 3:02 PM
To: thin@freelists.org
Subject: [THIN] Re: Citrix Load Evaluator / Total Sessions per Server & Application
Is the servername higher up the alphabet?
-----Original Message-----
From: thin-bounce@freelists.org [mailto:thin-bounce@freelists.org] On Behalf
Of TBarnhart@rcrh.org
Sent: 23 February 2009 21:34
To: thin@freelists.org
Subject: [THIN] Citrix Load Evaluator / Total Sessions per Server &
Application
Hey all,
We've had a small issue for the past year, and I can't seem to find an
answer about it.
We started a new 4.5 farm this time last year and have an application that
is published & load-managed across 20 servers.
There is a custom load-evaluator is built on these 4 "Assigned Rules":
Application User Load, Load Throttling, Memory Usage, Server User Load.
Overall, this load-evaluator appears to be working well for our userload;
per server and per application.
Except... One server seems to always have about 10-20% more total sessions
then the other 19 servers.
Over thousands of session - the 19 servers will all be within a range of 100
or so total sessions, but this one server will be always at the top of the
list for total sessions by several hundred more sessions for any time
period.
Theoretically - all of these are equal.
At anytime I run a "qfarm /app" (during average workday) - all of my servers
are in the 4500-6000 load range.
In the "Presentation Console" everything is running against this
load-evaluator.
Also, We have Citrix Edgesight collecting data, that looks normal other than
the total session count.
My performance statistics appear fairly normal at anytime we take a look.
The application is installed identically across the 20 servers.
Anyone have any ideas?
Thanks,
Troy
Troy Barnhart, Sr. Systems Programmer
tbarnhart@rcrh.org
Regional Health, Inc.
353 Fairmont Boulevard
Rapid City, South Dakota 57701
PH: 605-716-8352 / FAX: 605-716-8302
************************************************
For Archives, RSS, to Unsubscribe, Subscribe or
set Digest or Vacation mode use the below link:
http://www.freelists.org/list/thin
Follow ThinList on Twitter
http://twitter.com/thinlist
Thin List discussion is now available in blog format at:
http://thinmaillist.blogspot.com
Thinlist MOBILE Feed
http://thinlist.net/mobile
************************************************
************************************************
For Archives, RSS, to Unsubscribe, Subscribe or
set Digest or Vacation mode use the below link:
http://www.freelists.org/list/thin
Follow ThinList on Twitter
http://twitter.com/thinlist
Thin List discussion is now available in blog format at:
http://thinmaillist.blogspot.com
Thinlist MOBILE Feed
http://thinlist.net/mobile
************************************************
No comments:
Post a Comment