Hello everyone
We have a problem with mod_cluster and memory usage/sticky session:
If we set
KeepAlive On
KeepAliveTimeout 5
MaxKeepAliveRequests 0
globally in httpd.confmod_cluster looses the sticky session after a random amount of requests and routs a session to the wrong node.
As suggested in the minimalexample we then set
KeepAliveTimeout 60
MaxKeepAliveRequests 0
for the virtualhost that the JBoss servers connect to for their status messages. But now the httpd process uses more and more memory till it crashesthe server.
Is it requered to set
KeepAliveTimeout 60
MaxKeepAliveRequests 0
for mod_cluster to workcorrectly? I found no explanation in the documentation why it needs to be set like this.
Is there an other configuration that could explain the memory usage?
We use mod_cluster 1.1.0 on CentOS 5.4.
Thanks for your help,
Yves
The KeepAliveTimeout and MaxKeepAliveRequests values are there to prevent closing and reopening the connection from the nodes to httpd.
But it seems that there is a memory leak somewhere (MODCLUSTER-214).
Try MaxKeepAliveRequests 100.