1 Reply Latest reply on Mar 26, 2019 2:10 PM by rhusar

    Problem with detection of bad nodes in JBoss EAP 6.4 + Apache mod_cluster

    gian.honorio
    gian.honorio Mar 21, 2019 12:02 PM

      Hello, I have an environment that is running with JBoss EAP 6.4 and Apache mod_cluster, the JBoss environment contains a machine that is the domain responsible for controlling two nodes. below I'm putting my settings:

       

      http.conf

       

      <IfModule manager_module>

              Listen 192.168.0.100:8080

              ManagerBalancerName test-cluster-group

              <VirtualHost 192.168.0.100:8080>

                      <Location />

                              Require all granted

                      </Location>

                      TimeOut 30

                      KeepAliveTimeout 5

                      MaxKeepAliveRequests 0

                      AdvertiseFrequency 5

                      EnableMCPMReceive on

                      <Location /modcluster_manager>

                              SetHandler mod_cluster-manager

                              Order deny,allow

                              Allow from 192.168.0.0/16

                              Require all granted

                      </Location>

              </VirtualHost>

              <VirtualHost *:80>

                      ProxyPreserveHost On

                      ProxyPass /server-status !

                      ProxyPass / http://192.168.0.100:8080/app/

                      ProxyPassReverse / http://192.168.0.100:8080/app/

                      ProxyPassReverseCookiePath / /

                      <Location /server-status>

                              SetHandler server-status

                              Order deny,allow

                              Deny from all

                              Allow from 192.168.0.0/16

                      </Location>

              </VirtualHost>

      </IfModule>

       

      domain.xml

       

      <subsystem xmlns="urn:jboss:domain:modcluster:1.2">

           <mod-cluster-config advertise-socket="modcluster" proxy-list="192.168.0.100:8080" balancer="test-cluster-group" sticky-session="true" sticky-session-force="false" connector="ajp">

                <dynamic-load-provider>

                     <load-metric type="busyness"/>

                </dynamic-load-provider>

           </mod-cluster-config>

      </subsystem>

       

      My question is: How does Apache know that any of the nodes are malfunctioning? Is there any way to create some kind of personalized health check? Because in some cases from what I noticed, Apache continues to send requests to bad nodes, which ends up poisoning the cluster as a whole.