• Bug
  • Status: Open
  • 2 Major
  • Resolution:
  • prodmgmt
  • Reporter: kiwu
  • July 09, 2008
  • 0
  • Watchers: 0
  • March 19, 2010

Description

The JVM property tc.node-name changes the object name of the ClusterBean. This causes problems e.g for the ClusterBeanRegistrationListener that can be found in the jmx-utils project. This listener expects a cluster bean name of “org.terracotta:type=Terracotta Cluster,name=Terracotta Cluster Bean” and cannot find the bean if its name is something like “org.terracotta:clients=Clients,node-name=node1,type=Terracotta Cluster,name=Terracotta Cluster Bean”. Probably the node name property should not affect the object name of the cluster bean.

A more detailed discussion on this topic can be found here: http://forums.terracotta.org/forums/posts/list/1206.page

Comments

Steve Harris 2008-07-15

Can someone try to recreate this.

Fiona OShea 2008-07-23

Can you try to recreate this, and characterize the issue? thanks

Richard Dizmang 2008-09-08

2008-09-08 15:03:12,248 WARN - ************************************************************************************** The statistics buffer couldn’t be opened at ‘/Users/rdizmang/Terra/terracotta-2.7.0-nightly-rev10005-ee/samples/pojo/jtable/statistics-10.0.0.116’. The CVT system will not be active for this node.

A common reason for this is that you’re launching several Terracotta L1 clients on the same machine. The default directory for the statistics buffer uses the IP address of the machine that it runs on as the identifier. When several clients are being executed on the same machine, a typical solution to properly separate these directories is by using a JVM property at startup that is unique for each client.

For example: dso-java.sh -Dtc.node-name=node1 your.main.Class

You can then adapt the tc-config.xml file so that this JVM property is picked up when the statistics directory is configured by using %(tc.node-name) in the statistics path. **************************************************************************************

Just changing the error message to read -Dtc.node=node1 would make it much less likely that folks would be using -Dtc.node-name=node1. I would suggest that as a first fix.

Then, we should document the behavior of -Dtc.node-name=node1 as changing the bean name. This might well be useful in specific deployment situations.

Fiona OShea 2008-12-09

Taylor to have a meeting about Node names