[THIN] Re: Problems with MetaFram Presentation Server 3.0
- From: "Tim Mangan (new)" <tmangan@xxxxxxxxxxxx>
- To: <thin@xxxxxxxxxxxxx>
- Date: Mon, 21 Jun 2004 15:28:44 -0400
1) Why is there any load balancing going on if you have only one server?
2) 25,000 to 30,000 ctxsw/s is high but not necessarily excessive on that
hardware. Is that all the time or just when many are logging in (if the
latter, then printer mappings should be looked at).
_____
From: thin-bounce@xxxxxxxxxxxxx [mailto:thin-bounce@xxxxxxxxxxxxx] On Behalf
Of Friese, Casey A.
Sent: Monday, June 21, 2004 3:17 PM
To: thin@xxxxxxxxxxxxx
Subject: [THIN] Problems with MetaFram Presentation Server 3.0
Greetings, sorry for the cross post of this for those of you who have seen
this message in the NT List (thanks Roger for directing me here)
We just purchased and are attempting to put to use our first Citrix
environment. We are using the newest version of Citrix - MetaFrame
Presentation Server 3.0, which seems almost too new to find any good
resources about on the web.
Our configuration is as such (bear with me, I'm a Citrix novice so pardon
the misuse of terms)
We have Citrix installed on a Windows Server 2003 Enterprise edition OS
The server is a PowerEdge 2650 dual 3.0GHz processor with 4GB of RAM.
This server is the only server in the Citrix Farm
Despite reading that it isn't recommended to multi-home a Citrix server, we
have. One network card is facing the clients and the other is facing the
backend database used for the Application that the clients use.
The application that is being used is an "in house" application that is
written entirely in java. Currently 30 workstations are configured to run
this application from the Citrix server. Because each of the 30 users log
on to the network with the same credentials, the application has been
re-wrote to use the workstation name as the unique identifier when launched.
Normal execution of the application from the workstation pulls
%COMPUTERNAME% to get this unique identifier, through Citrix, the
application pulls %CLIENTNAME%. All of this works well unless 2 or more
stations connect at the exact same time. The result is that workstation A's
unique identifier shows correct and workstation B's shows the name of
workstation A. All of this works flawlessly in a plain Terminal Services
environment.
Secondly, the Citrix server is sending error events back because of High
Context Switching. From what I read, context switching in a Citrix
environment is one of the performance counters that determine the load on
the server. At around 15 concurrent connections the context switches hit
25,000-30,000 per second. I'm not sure if this is the cause of my next
issue or not so I'm not sure of 25 to 30k per second is high or low.
Thirdly, after about 15 concurrent connections my Citrix server stops
servicing connections from new clients. Upon firing off the ICA connection
for the public application, the Citrix server is contacted and then an error
message is displayed about a protocol error and the client can never
establish a connection to the application. Looking at traffic we're
noticing that the final ACK sent to the client by the Citrix server is
telling the client that the next "least busy" server is at a server that
doesn't even have Citrix installed and isn't even listening for TS or Citrix
connections.
By chance, can anyone offer some suggestions?
Thanks,
Casey Friese
Other related posts:
