[
https://issues.apache.org/jira/browse/FALCON-437?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13994254#comment-13994254
]
Satish Mittal commented on FALCON-437:
--------------------------------------
[~svenkat] Currently there are multiple copies of the same lib folder created
per feed - one copy within RETENTION folder, one copy for each source cluster
within REPLICATION. Wouldn't it be better if we upload just one copy of lib
folder and reuse it across retention and replication workflows?
> Feed Replication workflows are failing
> --------------------------------------
>
> Key: FALCON-437
> URL: https://issues.apache.org/jira/browse/FALCON-437
> Project: Falcon
> Issue Type: Bug
> Affects Versions: 0.5, 0.6
> Reporter: Satish Mittal
> Assignee: Venkatesh Seetharam
> Priority: Blocker
> Fix For: 0.5
>
> Attachments: FALCON-437.patch
>
>
> With latest trunk, feed replication workflows are failing with
> ClassNotFoundException for falcon lib classes (LateDataHandler and
> FalconPostProcessing).
> Noticed that in mapred.job.classpath.files conf property of replication jobs,
> none of falcon workflow lib jars are getting added. Hence the exception.
> Also noticed that when feed is scheduled, the lib folder is not getting
> created on HDFS at default appPath location.
> e.g. when replication appPath is:
> {noformat}
> ${nameNode}/projects/falcon/hcolo2/staging/falcon/workflows/feed/hcat-in-repl3/1399545050032/REPLICATION/hcat-cluster1
> {noformat}
> the lib folder is getting created at:
> {noformat}
> /projects/falcon/hcolo2/staging/falcon/workflows/feed/hcat-in-repl3/1399545050032/REPLICATION
> {noformat}
> It looks to be a regression due to FALCON-390.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
