Wildfly cluster and Apache2

[fpezzati]

Hello,

I'm running three Debian Wheezy guests on my host. Two of them run a Wildfly.8.2.0.Final cluster's node (IP are 192.168.0.101 and 192.168.0.102) the other guest run Apache2.2.22 with mod_cluster.1.2.0.Final module. This is my mod_cluster.conf:

<IfModlue mod_manager.so>
  Listen 192.168.0.104:6666
  ManagerBalancerName wfycluster
  <VirtualHost 192.168.0.104:6666>
  KeepAliveTimeout 300
  MaxKeepAliveRequests 0
  AdvertiseFrequency 5
  ServerAdvertise On 192.168.0.104:6666
  EnableMCPMReceive
  <Location /mod_cluster-manager>
  SetHandler mod_cluster-manager
  Order deny,allow
  Deny from all
  Allow from 192.168.0
  </Location>
  </VirtualHost>
</IfModule>

Modules are loaded correctly and I can see the mod_cluster's manager page on 192.168.0.104:6666/mod_cluster_manager but with no nodes info. I also configured a VirtualHost:

Listen 192.168.0.104:6666
<VirtualHost 192.168.0.104:6666>
  ServerName wfycluster
  ProxyPass / balancer://wfycluster
  ProxyPassReverse / balancer://wfycluster
  ProxyPreserveHost On 192.168.0.104
  <Location />
  Order deny,allow
  Allow from 192.168.0
  </Location>
  SetHandler mod_cluster-manager
  ManagerBalancerName other-server-group
  ErrorLog /var/log/apache2/wfycluster/error.log
</VirtualHost>

I run Wildfly istances by using the default standalone-ha.xml. This is the command:

./standalone.sh -b 192.168.0.101 -c standalone-ha.xml -Djboss.node.name=srv1 -u 230.0.0.4 -Djboss.bind.address.unsecure=192.168.0.101 -Djboss.bind.address.management=192.168.0.101

As nodes rises up there is no change in the Apache2 mod_cluster_manager page and I can see in virutal host's error log a lot of this messages:

...[error] proxy: CLUSTER: (balancer://wfycluster). All workers are in error state

The cluster is running correctly, I tested with curl directly on nodes.

[rhusar]

Modules are loaded correctly and I can see the mod_cluster's manager page on 192.168.0.104:6666/mod_cluster_manager but with no nodes info. I also configured a VirtualHost:

So even after you start the cluster, you can't see any nodes added to the manager page?

  ManagerBalancerName wfycluster

Most likely, did you also configure the same name for balancer in WildFly and reload?

[rhusar@x220 wildfly-8.1.0.Final]$ ./bin/jboss-cli.sh -c
[standalone@localhost:9990 /] /subsystem=modcluster/mod-cluster-config=configuration:write-attribute(name=balancer,value=wfycluster)
{
    "outcome" => "success",
    "response-headers" => {
        "operation-requires-reload" => true,
        "process-state" => "reload-required"
    }
}
[standalone@localhost:9990 /] :reload
{
    "outcome" => "success",
    "result" => undefined
}

Try adding a debug log to apache as well if that fails:

LogLevel debug

[fpezzati]

Hello,

I can't see no nodes in mod_cluster-manager page. I change my virtual host configuration this way:

Listen 192.168.0.104:6666

<VirtualHost 192.168.0.104:6666>

     ServerName wfycluster

     <Location />

     </Location>

     SetHandler mod_cluster-manager

     ManagerBalancerName wfycluster

     ServerAdvertise On 192.168.0.104:6666

     LogLevel debug

     ErrorLog /var/log/apache2/wfycluster/error.log

</VirtualHost>

I also add balancer=wfycluster attribute to mod-cluster-config tag in mod-cluster subsystem. Now Wildfly nodes report this error:

...(UndertowEventHandlerAdapter - 1) MODECLUSTER000042: Error null sending INFO command to debian1-2.local/192.168.0.104:6666, configuration will be reset: null

and Apache2's virtual host return this errors:

... [client 192.168.0.101] Invalid method in request INFO / HTTP/1.1

... [client 192.168.0.102] Invalid method in request INFO / HTTP/1.1

...

192.168.0.101 and 192.168.0.102 are Wildfly nodes ip addresses. Still no nodes on the mod_cluster-manager page. Maybe I'm doin' some mess with configuration.

[rhusar]

Listen 192.168.0.104:6666

<VirtualHost 192.168.0.104:6666>

    ServerName wfycluster

    <Location />

    </Location>

    SetHandler mod_cluster-manager

    ManagerBalancerName wfycluster

    ServerAdvertise On 192.168.0.104:6666

    LogLevel debug

    ErrorLog /var/log/apache2/wfycluster/error.log

</VirtualHost>

You are missing EnableMCPMReceive directive, thus this request is opaque and results in "Invalid method in request".

[fpezzati]

I was messing with Apache2 virtualhost configurations. I change my mod_cluster.conf this way:

<IfModule manager_module> 
 Listen 192.168.0.104:6666 
 ManagerBalancerName wfycluster 
 <VirtualHost 192.168.0.104:6666> 
 AllowDisplay On 
 KeepAliveTimeout 300 
 MaxKeepAliveRequests 0 
 AdvertiseFrequency 5 
 ServerAdvertise On 192.168.0.104:6666 
 AdvertiseGroup 224.0.1.105:23364 
 EnableMCPMReceive 
 <Location /> 
 Order allow,deny 
# Deny from all 
 Allow from 192.168.0. 
 </Location> 
 <Location /wfycluster> 
 SetHandler mod_cluster-manager 
 Order deny,allow 
 Allow from 192.168.0 
 </Location> 
 </VirtualHost> 
</IfModule> 

and change the other one virtualhost as simple as possible:

<VirtualHost *:80> 
 ServerName wfycluster 
 <Location /> 
 Order deny,allow 
 Allow from 192.168.0. 
 </Location> 
 LogLevel debug 
 ErrorLog /var/log/apache2/wfycluster/error.log 
</VirtualHost> 

Wildfly attribute balancer value and mod_cluster virtualhost ManagerBalancerName now matches. I can finally see nodes on mod_cluster-manager page. Maybe not the best conf possible, I have to dig deeper, but it works. Thanks for the help.

[rhusar]

Great.

Just make sure to remove the debug log level and consider adding advertise-security-key / advertiseSecurityKey configuration before going public with the setup.