[
https://issues.apache.org/jira/browse/HADOOP-12360?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14717117#comment-14717117
]
Steve Loughran commented on HADOOP-12360:
-----------------------------------------
One thing that's cropped up a couple of times is codahale metrics integration,
something which could be configured to push out its events to statds, graphite,
etc (though LGPL licensing would stop its ganglia plugin shipping with Hadoop)
Other than that, quick code review
# use NetUtils methods to get the hostname
# logging: you can use the SLF4J APIs and so skip the log level condition
wrappers; use debug() instead of trace() as log4j combines them anyway.
# you can use j7 <> types in constructors now
# exception catch clause in connect() should only catch {{IOException}} unless
there's some special reason otherwise
# inner statsd class should be private
The connection failure count logic isn't right for very-long lived apps; its
very different 5 failures in 30s from 5 failures over 20 days. Maybe after a
successful connect the counter should be reset.
very nice to see tests for this, but...
# the tests must have a timeout attribute so timing out operations (here:
receive()) have timeouts handled in the test runner, not a brute force jvm kill
# socket close() must be in finally clause, so that failed tests clean up
Finally, some documentation in the form of markdown will be needed to round off
the patch
> Create StatsD metrics2 sink
> ---------------------------
>
> Key: HADOOP-12360
> URL: https://issues.apache.org/jira/browse/HADOOP-12360
> Project: Hadoop Common
> Issue Type: New Feature
> Components: metrics
> Affects Versions: 2.7.1
> Reporter: Dave Marion
> Assignee: Dave Marion
> Priority: Minor
> Attachments: HADOOP-12360.001.patch, HADOOP-12360.002.patch
>
>
> Create a metrics sink that pushes to a StatsD daemon.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
