About Terracotta Documentation
This documentation is about Terracotta DSO, an advanced distributed-computing technology aimed at meeting special clustering requirements.
Terracotta products without the overhead and complexity of DSO meet the needs of almost all use cases and clustering requirements. To learn how to migrate from Terracotta DSO to standard Terracotta products, see Migrating From Terracotta DSO. To find documentation on non-DSO (standard) Terracotta products, see Terracotta Documentation. Terracotta release information, such as release notes and platform compatibility, is found in Product Information.
- How DSO Clustering Works
- Platform Concepts
- Hello Clustered World
- Setup and Configuration
- Planning for a Clustered App
- Configuring Terracotta DSO
- Configuration Reference
- Using Annotations
- Cluster Events
- Data Locality Methods
- Distributed Cache
- Clustered Async Data Processing
- Tool Guides
- Developer Console
- Operations Center
- tim-get (TIM Management Tool)
- Platform Statistics Recorder
- Eclipse Plugin
- Sessions Configurator
- Clustering Spring Webapp with Sessions Configurator
- Testing, Tuning, and Deployment
- Top 5 Tuning Tips
- Testing a Clustered App
- Tuning a Clustered App
- Deployment Guide
- Operations Guide
- FAQs and Troubleshooting
- General FAQ
- DSO Technical FAQ
- Troubleshooting Guide
- Non-portable Classes
- Migrating From DSO
- Concept and Architecture Guide
- Examinator Reference Application
- Clustered Data Structures Guide
- Integrating Terracotta DSO
- Clustering Spring Framework
- Integration Modules Manual
- AspectWerkz Pattern Language
The Terracotta server publishes a number of MBeans to provide information on the operation of the cluster. The Terracotta Console uses the JMX interface to provide real-time information, but the same interface can be used by any JMX client. As of Terracotta 2.3, there is also a new set of MBeans to report cluster events such as client nodes joining or leaving the cluster.
It is possible to configure the Terracotta Server to require authentication credentials before JMX communications take place. This will secure server shutdown, the Admin Console, JConsole, and any other processes attempting to connect via JMX. Server authentication will not provide secure message transmission once valid credentials are provided by a listening client. The JMX messages will not be encrypted. For this reason, we recommend that the Terracotta server be placed in a secure location on a private network and only be queried remotely via an encrypted tunnel such as those provided by stunnel or SSH.
Configuring JMX authentication with Terracotta requires that an authentication element be placed in the Terracotta configuration XML file, read by the Terracotta server instance that requires security. Terracotta relies on the standard Java security mechanisms for JMX which involve the creation of a .access and .password file with correct permissions set. The default location for these files is in (for JDK 1.5+) $JAVA_HOME/jre/lib/management.
Once configured properly, the Admin Console connection dialog will prompt you for credentials before completing a server connection. The stop-tc-server script located in the bin directory will also prompt for a username and password.
See the Configuration Guide and Reference for information on configuring the Terracotta server to use JMX authentication.
Setting permissions is done with the files jmxremote.password and jmxremote.access. Java security is very sensitive to the permission settings of these files so you will need to verify that they are properly set. One caveat to pay attention to is that the files must be owned by the system user who will be executing the Terracotta server. If the Terracotta server runs under the user "tcuser" then you must chown tcuser to change the ownership. The password file is most sensitive and requires:
NOTE: Setting these permissions on Windows is possible but requires a special DOS command. This permission is only available on Windows drives with NTFS formatting. See http://java.sun.com/j2se/1.5.0/docs/guide/management/security-windows.html for more information.
You may append to the default files located in $JAVA_HOME/jre/lib/management using the following commands:
'myusername' defines the desired username and 'mypassword' defines the desired password. In jmxremote.access, myusername is associated with the username defined in jmxremote.password and must declare readwrite permissions for that group.
The Terracotta Cluster Events JMX interface lets a Terracotta client listen for cluster events such as client connect and disconnect from the server and other clients connecting and disconnecting from the cluster.
As of Terracotta 2.3, when a Terracotta client connects to the cluster, it publishes a local MBean to the server called "Terracotta Cluster Bean". You can register with that bean to listen to cluster events. The published events are:
- com.tc.cluster.event.thisNodeConnected—The local client (the current JVM) connecting to a Terracotta server
- com.tc.cluster.event.thisNodeDisconnected—The local client disconnecting from a Terracotta server
- com.tc.cluster.event.nodeConnected—Another client connecting to a Terracotta server
- com.tc.cluster.event.nodeDisconnected—Another client disconnecting to a Terracotta server
You can also query the MBean for the local client's cluster node id (which is a unique id given to each client when it connects to the cluster for the first time) and for an enumeration of the node id's of each currently connected client.
Using this data, you can keep a local model of the Terracotta cluster. You can also find out if the local client has been disconnected from the cluster and take any further action.
Here's some sample code that will locate the local MBeanServer and register for cluster events:
To see a sample usage of the Terracotta Cluster Events interface, check out the jmx-sample (http://svn.terracotta.org/svn/forge/projects/jmx-sample) project from the Terracotta forge.
Then, start the Terracotta server. Next, run the demo.ClusterEvents class (in Eclipse as a Java Application or from the command line) with these VM arguments:
- Change /path/to/your/terracotta/installation to wherever you installed Terracotta
- Change dso-boot-hotspot_osx_150_07.jar to the appropriate boot jar for your platform
- You can also run dso-env.sh or dso-env.bat to determine the proper values for these properties
- The property -Dcom.sun.management.jmxremote lets you use jconsole to inspect the published MBeans.
You should see something like this:
You can then start another Terracotta client (one of the samples that comes with the Terracotta kit will work) and the demo.ClusterEvents program will print the client connect event to the Eclipse console. If you quit that client, the demo.ClusterEvents program will print the client disconnect event to the console:
If you stop the Terracotta server, the demo.ClusterEvents program will print a notification that this node has been disconnected:
If the Terracotta server is configured to be persistent, you can start the server up again and the demo.ClusterEvents program will print a notification that this node has been connected:
For another example of the Terracotta Cluster Events interface in use, see the chatter sample in the Terracotta kit.
For more information on the Terracotta cluster architecture, see the Concept and Architecture Guide
TODO: Write me. http://jira.terracotta.org/jira/browse/DOC-139
Rogue Client is a client which is for some reason not healthy e.g its having a lot of pending transactions or its slow.Terracotta server exposes a list of MBeans to monitor the performance of these clients (like pending transactions) so that one can determine the health of a client and can kill any particular unhealthy client.
Here is some sample code for getting client MBeans
These are the list of MBeans exposed to monitor these clients. The name of the APIs are self explanatory
- com.tc.stats.isTunneledBeansRegistered—Is the bean registered
- com.tc.stats.getL1InfoBeanName—Get the name(class ObjectName) of this bean
- com.tc.stats.getL1InfoBean—Get the MBean
- com.tc.stats.getInstrumentationLoggingBeanName—Get the name(class ObjectName) of the instrumentation logging MBean)
- com.tc.stats.getInstrumentationLoggingBean—Get the Instrumentation logging MBean
- com.tc.stats.getRuntimeLoggingBeanName—Get the name(Class ObjectName) of te runtime logging MBean
- com.tc.stats.getRuntimeLoggingBean—Get the run time logging MBean
- com.tc.stats.getRuntimeOutputOptionsBeanName—Get the name(class ObjectName) of the runtime output options MBean
- com.tc.stats.getRuntimeOutputOptionsBean—Get the runtime output options MBean
- com.tc.stats.getChannelID—Get the channel ID
- com.tc.stats.getRemoteAddress—Get the remote address of the channel
- com.tc.stats.getTransactionRate—Get the transaction rate for the client
- com.tc.stats.getObjectFaultRate—Get the object fault rate for the client
- com.tc.stats.getObjectFlushRate—Get the object flush rate for the client
- com.tc.stats.getPendingTransactionsCount—Get the pending transactions count for the client
- com.tc.stats.getStatistics—Get the statistics for the given properties
- com.tc.stats.killClient—Kill this client
TODO: Document the rest of the MBeans.