Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations Chris Miller on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Citrix load balancing out of step

Status
Not open for further replies.

kevsmart

MIS
Nov 29, 2001
4
GB
On a regular basis, we find load balancing going askew. We have 6 app servers in a load balanced farm and a separate master browser. On checking the browser stats via qserver /app we find one server (not always the same one!) reports a load of 10002 and this server seems to take all new logon connections. We have set Ibrowser logging but can't make any sense of what it reports.
Anyone help please?
Kev
 
if it is xp your running try looking at the server ica boardcast settings as they do have a habbit of changing
 
Sorry, I should have said, we are running NT4 t/s and MF 1.8 on all app servers. I have synchronised the time on all servers with the pdc, we have got the load balancing details at default settings, as all servers are identical. In the last 2 days we have seen 2 servers at a time taking more connections than others, i.e. 1 has 80 conns, another has 65, and the others are around 44. It's not always the same servers that do this. Does anyone know how to get Citrix to help with problems like this?
 
Load Balancing is a bit of a black art...

All your servers may be identical, but your user loads are different - use RMS or perfmon to find out what they're up to.

If you have particular users that regularly consume high levels of resources, eg with Java apps, maybe you could tie them to a particular server?

I hope this helps
 
Yes, I have checked perfmon. occasionally there are problems with spooler service hanging, possibly due to users having autocreated client printers getting disconnect while still having print jobs in the queue. I'm not convinced that this is a real issue but it may be a factor. I don't see how this or any users applications should affect load balancing which should be decided entirely on the no. of users per machine, or am I incorrect in that assumption.
I would like to know why the qserver info can be misleading - no. of connections rarely equals the number of users, what is it that constitutes a connection? - no. of connections + Free never equals no. of licences.
Any ideas?
 
Finally, Citrix have helped in solving this problem.

The problem was due to a number of thin client devices which had not been allocated client names. The way load balancing works is that when a client connection breaks without a clean disconnect the master browser caches its details. When that client reconnects, the browser sends it back to the server it was on before. Good eh? No, because any other clients with the same name get sent to the same server aswell, until the browser is reset. The solution is to ensure that all clients have unique names.
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top