[7u4] Request for approval for CR 7110104 - It should be possible to stop and start JMX Agent at runtime

[Dmitry Samersoff]

Problem

Currently, a JMX agent must be started from the Java command line in
order to enable JMX monitoring from a remote machine, though local
(same-machine) clients can attach dynamically. Once the agent is
started, it stops only when the JVM exits.

To diagnose problems in a production environment, we need the capability
to connect to a remote target JVM to verify or run queries, but that JVM
may or may not have a management agent started. We also need to be able
to stop a running agent in order to release the resources (e.g. opened
ports and memory) associated with the connection. The remote management
agent requires two TCP ports to communicate, one for the RMI registry
and other one for RMI itself. Currently we can specify only the RMI
registry port. The one for RMI is chosen randomly and not known to the
client. Despite the fact that most modern firewalls can handle this
situation with stateful rules, the ability to specify both ports is
desirable.


Solution

Create diagnostic commands usable by the jcmd utility to stop running a
jmx agent and then start it again with different properties. The syntax
is similar to the the way the information is specified on the Java
command line.


JDK8 changeset:

 http://hg.openjdk.java.net/hsx/hsx23/hotspot/rev/3322dd7f0a4d
 http://hg.openjdk.java.net/jdk8/tl/jdk/rev/1e737abbff6f


List of reviewers:

Mandy Chung
Karen Kinnear


-- 
Dmitry Samersoff
Java Hotspot development team, SPB04
* There will come soft rains ...