Okay, here it goes (sorry for the length of this post ( :| ):
We have three seperate servers in our setup:
192.168.0.1 webserver
192.168.0.2 appserver1
192.168.0.3 appserver2
We configured the DNS such that www.application1.nl and www.application2.nl are landing on the webserver. The aim is to get application contexts /application1 and /application2 mapped to there own cluster:
cluster1 contains JBoss instances application1_i1, application1_i2, running on respectively
appserver1 and appserver2, and reacheable via http://appserver1:8009/application1
cluster2 contains JBoss instances application2_i1, application2_i2, running on respectively
appserver1 and appserver2, and reacheable via http://appserver2:8109/application2
-------------------------------------------------------------------------------------------
On webserver, we have the following configuration items:
In the first conf (placed in the conf.d directory) the modules needed for mod_cluster are
loaded into httpd:
$ cat mod_cluster.conf
LoadModule proxy_module modules/mod_proxy.so
LoadModule proxy_ajp_module modules/mod_proxy_ajp.so
LoadModule slotmem_module modules/mod_slotmem.so
LoadModule manager_module modules/mod_manager.so
LoadModule proxy_cluster_module modules/mod_proxy_cluster.so
LoadModule advertise_module modules/mod_advertise.so
At the end of the httpd.conf, we load the conf files of the virtual host. We do this to
establish an order (naming them 0.conf etc works in the same way):
$ tail httpd.conf
Include "conf/virtual_hosts/jboss.conf"
Include "conf/virtual_hosts/default.conf"
Include "conf/virtual_hosts/application1conf"
Include "conf/virtual_hosts/application2.conf"
The vitrual host confs are the following:
$ cat jboss.conf
Listen 192.168.0.1:7777
<VirtualHost 192.168.0.1:7777>
KeepAliveTimeout 60
MaxKeepAliveRequests 0
ManagerBalancerName mycluster
AdvertiseFrequency 5
ServerAdvertise Off
Createbalancers 1
</VirtualHost>
$ cat default.conf
<VirtualHost *:80>
ServerAdmin unix@tntpost.nl
DocumentRoot /apps/nlptc27b14/data
ServerName nlptc27b14.nlptc27.post.tnt:80
ErrorLog logs/nlptc27b14-error_log
CustomLog logs/nlptc27b14-access_log combined
LogLevel debug
# mod_cluster defaults
<Location /mcm>
SetHandler mod_cluster-manager
</Location>
</VirtualHost>
$ cat application1.nl.post.tnt.conf
<VirtualHost *:80>
ServerAdmin unix@tntpost.nl
DocumentRoot /apps/www.application1.nl/data
ServerName www.application1.nl:80
ErrorLog logs/www.application1.nl-error_log
CustomLog logs/www.application1.nl-access_log combined
LogLevel debug
ProxyPass / balancer://cluster1/application1 stickysession=JSESSIONID|jsessionid
nofailover=On
ProxyPassReverse / balancer://cluster1/application1
</VirtualHost>
$ cat application2.nl.post.tnt.conf
<VirtualHost *:80>
# Begin rewriting TRACK/TRACE
RewriteEngine on
RewriteCond %{REQUEST_METHOD} ^TRACK
RewriteRule .* - [F]
RewriteCond %{REQUEST_METHOD} ^TRACE
RewriteRule .* - [F]
# End rewriting TRACK/TRACE
ServerAdmin unix@tntpost.nl
DocumentRoot /apps/www.application2.nl/data
ServerName www.application2.nl:80
ErrorLog logs/www.application2.nl-error_log
CustomLog logs/www.application2.nl-access_log combined
ProxyPass / balancer://cluster2/application2 stickysession=JSESSIONID|jsessionid
nofailover=On
ProxyPassReverse / balancer://cluster2/
</VirtualHost>
-----------------------------------------------------------------------------------------
On appserver1. we have (for example) instance application1_i1, which is primarily a clone of the production instance (on Redhat JBoss 5, this is the same as a clone of the all instance). The following items are changed:
1. mod-cluster.sar is deployed in application1_i1.
2. application1 is deployed in application1_i1.
3. In the mod-cluster.sar/META-INF/mod-cluster-jboss-beans.xml we added the following
propeties to the "HAModClusterConfig" and the "ModClusterService":
<property name="balancer">cluster1</property>
<property name="proxyList">${jboss.modcluster.proxyList:}</property>
The variable ${jboss.modcluster.proxyList:} is given during startup via a -D option.
4. In the jbossweb.sar/server.xml we added the following things:
a. Added the following listener:
<Listener
className="org.jboss.web.tomcat.service.deployers.MicrocontainerIntegrationLifecycleListener" delegateBeanName="HAModClusterService"/>
b. added jvmRoute="${jboss.jvmRoute}" to the jboss.web Engine entry
The variable ${jboss.jvmRoute} is given during startup via a -D option.
--------------------------------------------------------------------------------------------
On appserver1, we have instance application2_i1, which is the same as application1_i1. The only difference with application1_i2 is that we deploy application2 in this instance.
--------------------------------------------------------------------------------------------
On appserver2 we have the same configuration, only now that instances are called application1_i2 and application2_i2
--------------------------------------------------------------------------------------------
On appserver1, we can now start (for example) instance application1_i1 as follows:
cd /opt/jboss/jboss-as/bin; /opt/jboss/jboss-as/bin/run.sh -c application1_i1 -b appserver1 -g
cluster1 -u 239.255.100.100 -m 60100 -Djboss.messaging.ServerPeerID=0
-Djboss.service.binding.set=ports-default -Djboss.server.log.dir=/apps/data/test/fbewbrpp/logs/
-Djboss.jvmRoute="appserver1" -Djboss.Domain=cluster1
-Djboss.modcluster.proxyList="192.168.0.1:7777"
On appserver2, we can now start (for example) instance application1_i2 as follows:
cd /opt/jboss/jboss-as/bin; /opt/jboss/jboss-as/bin/run.sh -c application1_i2 -b appserver2 -g
cluster1 -u 239.255.100.100 -m 60100 -Djboss.messaging.ServerPeerID=1
-Djboss.service.binding.set=ports-default -Djboss.server.log.dir=/apps/data/test/fbewbrpp/logs/
-Djboss.jvmRoute="appserver1" -Djboss.Domain=cluster1
-Djboss.modcluster.proxyList="192.168.0.1:7777"
Now these form one cluster. We can do the same with cluster2 = (application2_i1,
application2_i2).
-------------------------------------------------------------------------------------------
Problem is still: When I run the configuration with 1 apache virtual host (www.application1.nl), then www.application1.nl/application1 is available and the cluster manager at http://webserver/mcm give two active nodes.
When I run the configuration with 2 apache virtual hosts (www.application1.nl and www.application2.nl), then www.application1.nl/application1 is unavailable (Temporary unavailable while lynx http://appserver1:8080/application1 from the webserver gives the application1), and
www.application2.nl/application2 is available. In this case I see two nodes in the cluster manager (I expect 4) and not all the information there can I call reliable.