[
https://issues.apache.org/jira/browse/HADOOP-7436?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13058051#comment-13058051
]
Eric Yang commented on HADOOP-7436:
-----------------------------------
Metrics plugin is like a driver software. Linux kernel bundles compatible
drivers in the source tarball. This is no different, it is easier to certify a
software driver if it is included in the kernel source with test cases. If a
broken driver is not fixed, it is easy to remove. The power and
responsibilities are in the hand of the Hadoop PMC. If Hadoop community don't
make integration easy, then it depends on third party to make the integration
happen. The distributed integration approach only drains developers from
Hadoop community. This is something I would like to avoid.
> Bundle Chukwa Metrics plugin in Hadoop
> --------------------------------------
>
> Key: HADOOP-7436
> URL: https://issues.apache.org/jira/browse/HADOOP-7436
> Project: Hadoop Common
> Issue Type: New Feature
> Environment: Java 6
> Reporter: Eric Yang
> Assignee: Eric Yang
>
> For monitoring hadoop cluster with Chukwa, the current step is to copy
> chukwa-hadoop-*-client.jar and json-simple to hadoop classpath. (i.e.
> $HADOOP_HOME/lib or $HADOOP_PREFIX/share/hadoop/lib), and modify the
> hadoop-metrics.properties to use
> org.apache.hadoop.chukwa.inputtools.log4j.Log4JMetricsContext for emitting
> metrics. It is preferred to reduce the number of manual steps that is
> required to enable chukwa monitored hadoop cluster by moving the plugin code
> into hadoop code base. It is similar to bundling Ganglia metrics plugin in
> Hadoop code base.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
