chukwa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Fay Wang (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CHUKWA-819) Add CaffeOnSpark to Chukwa docker
Date Mon, 01 May 2017 14:56:04 GMT

    [ https://issues.apache.org/jira/browse/CHUKWA-819?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15990916#comment-15990916
] 

Fay Wang commented on CHUKWA-819:
---------------------------------

For distributed system, diagnose service and application problems can be challenging sometimes.
 The main challenge is that the information are scattered across nodes and services.  The
amount of data involved to describe the problem becomes non-trivial.  For simple problems,
it can be explained by mapping job and task logs to isolate a particular problem.  More complex
problems will involve trace through Spark, YARN, HBase and HDFS log files.  This feature describes
mechanism that can be developed to support debugging more complex problems by using computer
vision and machine learning algorithm to filter out noise and identify similar patterns to
find the root cause of problems.

> Add CaffeOnSpark to Chukwa docker 
> ----------------------------------
>
>                 Key: CHUKWA-819
>                 URL: https://issues.apache.org/jira/browse/CHUKWA-819
>             Project: Chukwa
>          Issue Type: New Feature
>    Affects Versions: 0.8.0
>            Reporter: Fay Wang
>             Fix For: 0.9.0
>
>
> Add CaffeOnSpark to Chukwa docker image to support memory leak image machine training
using Caffe.  



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Mime
View raw message