[
https://issues.apache.org/jira/browse/PIG-4054?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Daniel Dai updated PIG-4054:
----------------------------
Description:
Currently Pig creates job.jar per job when submitting mapreduce job. There are
several disadvantages:
1. job.jar varies job by job, job.jar will not get reused even if jar cache is
used (PIG-2672).
2. Before job submission, we need to pack a job.jar which are mostly repacking
of existing jars, this is a waste of time
3. job.jar is a uber jar which makes debug harder and could lead to jar
conflicting issue (eg, PIG-3039)
On tez side, situation is similar, the consequence is worse since container
will not be reused.
So instead of job.jar, I would like to ship individual jar to distributed
cache. Note this issue is in essence independent of PIG-4047, however, PIG-4047
would make the picture more complete in that we don't have any uber jars.
was:
Currently Pig creates job.jar per job when submitting mapreduce job. There are
several disadvantages:
1. job.jar varies job by job, job.jar will not get reused if jar cache is used
(PIG-2672).
2. Before job submission, we need to pack a job.jar which are mostly the same,
this is a waste of time
3. job.jar is a uber jar which makes debug harder and could lead to jar
conflicting issue (eg, PIG-3039)
On tez side, situation is similar, the consequence is worse since container
will not be reused.
So instead of job.jar, I would like to ship individual jar to distributed
cache. Note this issue is in essence independent of PIG-4047, however, PIG-4047
would make the picture more complete in that we don't have any uber jars.
> Do not create job.jar when submitting job
> -----------------------------------------
>
> Key: PIG-4054
> URL: https://issues.apache.org/jira/browse/PIG-4054
> Project: Pig
> Issue Type: Improvement
> Components: impl
> Reporter: Daniel Dai
> Assignee: Daniel Dai
> Fix For: 0.14.0
>
> Attachments: PIG-4054-1.patch
>
>
> Currently Pig creates job.jar per job when submitting mapreduce job. There
> are several disadvantages:
> 1. job.jar varies job by job, job.jar will not get reused even if jar cache
> is used (PIG-2672).
> 2. Before job submission, we need to pack a job.jar which are mostly
> repacking of existing jars, this is a waste of time
> 3. job.jar is a uber jar which makes debug harder and could lead to jar
> conflicting issue (eg, PIG-3039)
> On tez side, situation is similar, the consequence is worse since container
> will not be reused.
> So instead of job.jar, I would like to ship individual jar to distributed
> cache. Note this issue is in essence independent of PIG-4047, however,
> PIG-4047 would make the picture more complete in that we don't have any uber
> jars.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
