Plugin doest respond until web server reboot

Question

We have a weblogic 6.0SP2 cluster with 4 backends and two Iplanet Web Server 4.1 as
front ends.
During FULL GCs in weblogic JVM cluster nodes do not respond to any request and the
foront ends are not able to respond.
Web have noticed that if the error persists serveral seconds (or minutes) and then
weblogic recovers the web server cannot conect anymore to webloguc until we reboot
it.
This is what Iplanets error log says:
[XX/Jun/2002:15:52:10] failure ( 628): for host xx.xx.xx.xx trying to GET /some_weblogic_URL,
wl-proxy reports: exception occurred for backend host 'my_cluster_node/my_port':
'READ_TIMEOUT [os error=22,  line 151 of Reader.cpp]: no read after 300 seconds'
Has anyone else experience this? how can we solve it?
PLEASE REPLY!
TIA
Cristina Ceballos

Answer 1

in addition to, To simulate the problem, you need to set your window registry to allow IE to open up 20 connections (default is 2) per IE. Please follow the steps below :
1. run "regedit" to open up the registry
2. select HKEY_CURRENT_USER|Software|Microsoft|windows|CurrentVersion|Internet Settings
3. right click on the right hand panel and select New|DWORD Value
4. type "MaxConnectionsPer1_0Server" for the name
5. right click on the newly created "MaxConnectionsPer1_0Server" key and select "Modify"
6. enter 32 as hexadecimal value in the text box then click OK button to save
7. repeat step 3 to 6 with the name as "MaxConnectionsPerServer"
8. close all your IE browsers
if you do not do the above settings, you need to open up 30 browsers to simulate 60 connections to web server.
Once the a.jsp exit from the while loop, you should see the same number of requests in your iplanet access log. And, when you view source of a frame, you should see "<br>count=1<br>count=2<br>count=3<br>count=4<br>count=5"; otherwise the request is not handled by web server.