[
https://issues.apache.org/jira/browse/HADOOP-6350?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14025978#comment-14025978
]
Arpit Agarwal commented on HADOOP-6350:
---------------------------------------
I reviewed the HDFS/common portions. A few comments below. Nothing worth
blocking the checkin as this is a huge improvement over no documentation.
# I am not sure what this means - _Each metrics record contains tags such as
ProcessName, SessionId, and Hostname as additional information along with
metrics._. How are these tags accessed, I don't see them in jconsole? Perhaps I
am missing some basic knowledge, let me know if so.
# Namenode - snapshot metrics are missing.
# DataNode - DataNodeInfo metrics are missing.
# DataNode - FsDatasetState metrics are missing.
# Nitpick: we should use title case consistently for sub-headings e.g.
rpcdetail --> RpcDetailed
As a separate discussion I think long term maintenance of this documentation
will be challenging. Does Java support declarative definition of metrics which
can be used to auto-generate this documentation?
Thanks for taking up this work!
> Documenting Hadoop metrics
> --------------------------
>
> Key: HADOOP-6350
> URL: https://issues.apache.org/jira/browse/HADOOP-6350
> Project: Hadoop Common
> Issue Type: Improvement
> Components: documentation, metrics
> Affects Versions: 3.0.0, 2.1.0-beta
> Reporter: Hong Tang
> Assignee: Akira AJISAKA
> Labels: metrics
> Attachments: HADOOP-6350-sample-1.patch, HADOOP-6350-sample-2.patch,
> HADOOP-6350-sample-3.patch, HADOOP-6350.4.patch, HADOOP-6350.5.patch,
> HADOOP-6350.6.patch, HADOOP-6350.7.patch, sample1.png
>
>
> Metrics should be part of public API, and should be clearly documented
> similar to HADOOP-5073, so that we can reliably build tools on top of them.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
