[
https://issues.apache.org/jira/browse/YARN-7265?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16190285#comment-16190285
]
Eric Yang commented on YARN-7265:
---------------------------------
Chukwa has some degree of success in doing this. Chukwa 0.8.0 utilize SOLR to
provide indexing feature. This allows user to search for unique text (i.e.
Application ID), and all logs contains unique text are categorized and display
as results. This includes user's container logs, hdfs (nn/dn), yarn (rm/nm) ,
and hbase (master/rs) log files. This exists as proof of concept. Chukwa
docker image is available on Docker hub for testing. There is a lot work to
productize it, but feel free to experiment with it.
> Hadoop Server Log Correlation
> -------------------------------
>
> Key: YARN-7265
> URL: https://issues.apache.org/jira/browse/YARN-7265
> Project: Hadoop YARN
> Issue Type: Wish
> Components: log-aggregation
> Reporter: Tanping Wang
>
> Hadoop has many server logs, yarn tasks logs, node manager logs, HDFS logs..
> There are also a lot of different ways can be used to expose the logs, build
> relationship horizontally to correlate the logs or search the logs by
> keyword. There is a need for a default and yet convenient logging analytics
> mechanism in Hadoop itself that at least covers all the server logs of
> Hadoop. This log analytics system can correlate the Hadoop server logs by
> grouping them by various dimensions including application ID, task ID, job
> ID or node ID etc. The raw logs with correlation can be easily accessed by
> the application developer or cluster administrator via web page for managing
> and debugging.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
