[jboss-dev-forums] [Design of Management Features on JBoss] - Node identification

[bstansberry@jboss.com]

Tim Fox and I were having a discussion on http://jira.jboss.com/jira/browse/JBMESSAGING-957 about how to give identity to servers in a cluster.  Wanted to bring it here to get inputs from others who may face similar issues and in particular from the JON folks.

Issue is basically that there are various places where code needs to give some kind of identity to a server:

1) Tomcat engine if mod_jk is used (jvmRoute attribute).
2) JBoss Messaging if clustered.
3) HAPartition to prevent two incarnations of the same server joining the cluster.

Tim and I are discussing setting the name for a server via run.sh, perhaps using something like what's done with $JBOSS_HOST and -b in jboss_init_redhat.sh:

JBOSS_NODE_NAME=${JBOSS_NODE_ID:+"-i $JBOSS_NODE_ID"}
  | #define the script to use to start jboss
  | JBOSSSH=${JBOSSSH:-"$JBOSS_HOME/bin/run.sh -c $JBOSS_CONF $JBOSS_BIND_ADDR $JBOSS_NODE_ID"}

The above presumes the addition of a new startup switch '-i' or --id; a system property could be used as well.

Basically just wanted to get any comments any one had adding this sort of thing, e.g.

1) what are your use cases for wanting an id for a server?
2) does it need to be persistent across restarts (i.e. can't be randomly generated at runtime or something)?
3) Any constraints on the id? E.g. JBM wants an int id, which isn't really appropriate for the jvmRoute or HAPartition use cases.
4) Any thoughts on setting it based on the local env via the startup script?  Would the JON folks want to do it some other way?


View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4045578#4045578

Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4045578