{"id":1238,"date":"2012-08-30T16:14:55","date_gmt":"2012-08-30T08:14:55","guid":{"rendered":"http:\/\/rmohan.com\/?p=1238"},"modified":"2012-08-30T16:16:00","modified_gmt":"2012-08-30T08:16:00","slug":"tomcat-clustering-and-failover","status":"publish","type":"post","link":"https:\/\/mohan.sg\/?p=1238","title":{"rendered":"Tomcat clustering and failover"},"content":{"rendered":"

\u00a0<\/h2>\n

In this post I will describe how to implement a functional Tomcat cluster with load balancing and failover<\/p>\n

I did a great deal of research trying to get this environment working. There are a number of good tutorials for simple Tomcat clustering and load balancing. I was unable to find an example that was complete which provided a load balanced solution with failover.<\/p>\n

Example 1 shows load balancing with a single server.<\/p>\n

\"\"<\/a><\/p>\n

Example 2 shows load balancing with a multiple Tomcat servers.<\/p>\n

\"\"<\/a><\/p>\n

Example 3 shows load balancing with failover which is described in this post.<\/p>\n

\"\"<\/a><\/p>\n

The first step in setting up a cluster like we have above is to insure that multicast works in your environment. In order to check for multicast support on Linux first check to see if your kernel was compiled with multicast support by trying the following:<\/p>\n

cat proc\/net\/dev_mcast<\/code><\/p>\n

If the file doesn\u2019t exist a kernel recompile is in order, the instructions for this are outside the scope of this post. Assuming we have kernel multicast support we must check that it is working. This can be done with a simple java utility contained in the zipfile\u00a0multicasttest.zip<\/a>. There are 2 classes in the zipfile,\u00a0TestServer and\u00a0TestClient first start the test client on all nodes in the cluster, then run TestServer on one of the nodes with the number of messages to send as a count. This will send a message every 3 seconds until the message count is reached. All of the clients should report each message in the form \u201cReceived: Message 1 of 10? and the server will display \u201cSending message 1?. If each of the clients report the messages your clustering should work.<\/p>\n

Next configure your Tomcat servers for clustering and failover. Samples of the configuration file shown below are contained in the zipfile SampleConfig.zip<\/a>.<\/p>\n

Add the following block in your [TOMCAT]\/conf\/server.xml:<\/p>\n

 <\/p>\n

 <\/p>\n

<!\u2013 mod_jk ajp protocol connector \u2013><\/p>\n

<Connector port=\u201d8009? protocol=\u201dAJP\/1.3? redirectPort=\u201d8443? server=\u201dApache\u201d \/><\/p>\n

 <\/p>\n

<!-- Engine with clustering support set jvmRoute to \"jvm\"+id unique to tomcat instance --><\/code><\/pre>\n
 <\/p>\n
<Engine name=\"Catalina\" defaultHost=\"localhost\" jvmRoute=\"jvm2\"><\/code><\/pre>\n
 <\/p>\n
<Cluster className=\"org.apache.catalina.ha.tcp.SimpleTcpCluster\" channelSendOptions=\"8\"><\/code><\/pre>\n
 <\/p>\n
<Manager className=\"org.apache.catalina.ha.session.DeltaManager\" expireSessionsOnShutdown=\"false\" notifyListenersOnReplication=\"true\"\/><\/code><\/pre>\n
 <\/p>\n
 <\/p>\n
 <\/p>\n
 <\/p>\n
<Channel className=\"org.apache.catalina.tribes.group.GroupChannel\"><\/pre>\n
<!--Membership address and port should be set to the proper<\/pre>\n
value for the environment --><\/pre>\n
<Membership className=\"org.apache.catalina.tribes.membership.McastService\"<\/pre>\n
address=\"228.0.0.4\" port=\"45564\" frequency=\"500\" dropTime=\"3000\"\/><\/pre>\n
<Sender className=\"org.apache.catalina.tribes.transport.ReplicationTransmitter\"><\/pre>\n
<Transport className=\"org.apache.catalina.tribes.transport.nio.PooledParallelSender\"\/><\/pre>\n
<\/Sender><\/pre>\n
<Receiver className=\"org.apache.catalina.tribes.transport.nio.NioReceiver\"\r\naddress=\"auto\" port=\"4000\" autoBind=\"100\" selectorTimeout=\"5000\" maxThreads=\"6\"\/><\/pre>\n
<Interceptor className=\"org.apache.catalina.tribes.group.interceptors.TcpFailureDetector\"\/><\/pre>\n
<Interceptor className=\"org.apache.catalina.tribes.group.interceptors.MessageDispatch15Interceptor\"\/><\/pre>\n
<\/Channel>\r\n<Valve className=\"org.apache.catalina.ha.tcp.ReplicationValve\" filter=\"\"\/>\r\n<Valve className=\"org.apache.catalina.ha.session.JvmRouteBinderValve\"\/>\r\n<ClusterListener className=\"org.apache.catalina.ha.session.JvmRouteSessionIDBinderListener\"\/>\r\n<ClusterListener className=\"org.apache.catalina.ha.session.ClusterSessionListener\"\/>\r\n<\/Cluster><\/pre>\n
 <\/p>\n
 <\/p>\n
 <\/p>\n
 <\/p>\n
Create an [APACHE]\/conf\/httpd.conf:
 Include conf\/mod_jk.conf<\/p>\n
Save the following to [APACHE]\/conf\/mod_jk.conf and enter the following:<\/p>\n
LoadModule jk_module modules\/mod_jk.so<\/p>\n
# Specify path to worker configuration file
 JkWorkersFile conf\/workers.properties<\/p>\n
# Configure logging and memory
 JkShmFile logs\/mod_jk.shm
 JkLogFile logs\/mod_jk.log
 JkLogLevel debug<\/p>\n
# Configure monitoring
 JkMount \/jkmanager\/* jkstatus<\/p>\n
<Location \/jkmanager>
 Order deny,allow
 Deny from all
 Allow from localhost
 <\/Location><\/p>\n
# Configure applications
 JkMount \/myApp\/* loadbalancer
 JkMount \/myOtherApp\/* loadbalancer<\/p>\n
 <\/p>\n
Create an [APACHE]\/conf\/workers.properties and enter the following:<\/p>\n
# Define worker names
 worker.list=worker1, worker2, jkstatus, loadbalancer<\/p>\n
# Create virtual workers
 worker.jkstatus.type=status
 worker.loadbalancer.type=lb<\/p>\n
# Declare Tomcat server worker 1 on the localhost
 worker.worker1.type=ajp13
 worker.worker1.distance=1
 worker.worker1.port=8009
 worker.worker1.host=localhost<\/p>\n
# Declare Tomcat server worker 2 on the remote host
 # setting the host value to the remote ip address
 worker.worker2.type=ajp13
 worker.worker2.distance=10
 worker.worker2.port=8009
 worker.worker2.host=XXX.XXX.XXX.XXX #IP address of remote server<\/p>\n
# Associate real workers with virtual LoadBalancer worker
 worker.loadbalancer.balanced_workers=worker1, worker2
 worker.loadbalancer.sticky_session=1<\/p>\n
You should now be able to start your Apache and Tomcat instances and have clustering and failover.<\/p>\n
For this example Tomcat 6.0.29 was used with Apache 2.2 and mod_jk 1.2.31.<\/p>\n
 <\/p>\n
 <\/p>\n
 <\/p>\n
Clustering allows us to run an application on several parallel servers (i.e. cluster nodes). The load is distributed across different servers, and even if any of the servers fails the application is still accessible via other cluster nodes. Clustering is crucial for scalable enterprise applications, as you can improve performance by simply adding more nodes to the cluster.<\/p>\n
This document provides step by step instructions for configuring Apache server which handles\u00a0\u00a0 
request for static contents and delegates request for dynamic pages (i.e. JSP or Servlet) to two\u00a0 instances of Tomcat server using JK Connectors.<\/p>\n