We have a small web cluster for servering about max 400 concurrent user each day.
We using :
- Apache 2.2.14
- JBoss 4.2.2 GA
Firstly we tried AJP and unforutanly we got AJP
CLOSE_WAIT problem which caused apache not response to HTTP request. So we turn to
mod_proxy_balancer, it looks not have the same
CLOSE_WAIT problem in each JBoss node.
Here is our configuration for mod_proxy_balancer now:
<Proxy balancer://www> Allow from all BalancerMember http://192.168.1.2:8080/APP loadfactor=10 route=node1 BalancerMember http://192.168.1.3:8080/APP loadfactor=10 route=node2 BalancerMember .... loadfactor=10 route=node3 ProxySet lbmethod=byrequests stickysession=JSESSIONID|jsessionid nofailover=On </Proxy> ProxyPass /LEI balancer://www/ ProxyPassReverse /LEI balancer://www/
But we got a new problem here, there were many (hundreds of ) CLOSE_WAIT we can find from the apache node machine:
netstat -a | grep CLOSE_WAIT ............. tcp 1 0 localhost:30997 localhost:webcache CLOSE_WAIT
Is this a problem of
mod_proxy_balancer? How can we avoid this?
We have read all post about
CLOSE_WAIT in this forum and tried many solutions when we using AJP.
Are they any other http cluster solutions without using Apache
BTW: We have read nginx and it only support balance based on client IP router, right? Most of our users are inside one large intranet (campus users).
Thanks for any help.