Christian Kunz wrote:
Is there support to tell hadoop servers (tasktracker, datanode, ....)
to re-read configuration, or fully reset, or to shut down (without an
external kill)?
Some of the lifecycle stuff that Is now targeted at 0.21 will be able to
do this.
https://issues.apache.org/jira/browse/HADOOP-3628
I will put up a new patch soon (in sync with SVN_HEAD), and would
welcome more work integrating it with different management tools. To
propagate configuration changes around you also need to get the
configuration in from places other than XML files on the classpath. i've
retrofitted exactly that into our own code by subclassing JobConf, but
something cleaner could be done, something that would be directly driven
off LDAP or other databases as well as SmartFrog
-steve
