tajo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jihoon Son (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (TAJO-333) Add metric system to Tajo
Date Tue, 10 Dec 2013 06:16:07 GMT

    [ https://issues.apache.org/jira/browse/TAJO-333?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13843992#comment-13843992
] 

Jihoon Son commented on TAJO-333:
---------------------------------

In overall, the patch looks good to me.
But, I have a couple of things to discuss.
* For the high readability, it will be better to change the names of metrics to be more meaningful.
For example, "executedExecutionBlocksNum" looks more meaningful than "runEB". Although the
name lengths will increase, but it can increase readability.
* The default name of metrics file can be better to move from TajoSystemMetrics to TajoConf.ConfVars.
Also, it will be better to use TajoConf.getVar() instead of TajoConf.get() in TajoSystemMetrics.constructor().
* In TajoSystemMetrics, propertyChangeChecker seems to have the same role with FileChangedReloadingStrategy.
If this is my misunderstading, please explain to me.
* Please remove the unused main function from TajoSystemMetrics.
* In MetricsConsoleReporter, the hostAndPort is fixed as "localhost" and it looks to represent
the master address. But, I think that it may be different in some environments. So, how about
read the address from QueryMaster?
* Please remove a debug code at Line 269 in TajoWorker.
* In tajo-metrics.propertis, there is a typing error at Line 46. (worker-jvm-reporters=cnull
=> worker-jvm-reporters=null)
* Please remove unused imports.

And, how about add a metric for the size of intermediate data?
I think that it will be useful.

> Add metric system to Tajo
> -------------------------
>
>                 Key: TAJO-333
>                 URL: https://issues.apache.org/jira/browse/TAJO-333
>             Project: Tajo
>          Issue Type: New Feature
>            Reporter: Hyunsik Choi
>            Assignee: hyoungjunkim
>              Labels: metrics
>             Fix For: 0.8-incubating
>
>         Attachments: TAJO-333.patch
>
>
> Currently, there is no way to measure and observe Tajo's internal operations. As a Hadoop
job, Tajo also should measure and record a lot of operational works.
> I would like to propose the use of Metrics (http://metrics.codahale.com/). It is well
known as a stable and efficient metric library.



--
This message was sent by Atlassian JIRA
(v6.1.4#6159)

Mime
View raw message