[
https://issues.apache.org/jira/browse/TEZ-1247?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14056944#comment-14056944
]
Bikas Saha commented on TEZ-1247:
---------------------------------
bq. I plan to add a new API in TezClient ( eg. explainDAG ) for generating dot
file ( may be other stuff in future ), still using submitDAG may be a little
confusing to users
I usually not a good idea to mix API's for different things on the same object.
Secondly, it does not address all scenarios. Its likely that we need to debug a
hive job thats not working correctly and so we need to debug the DAG plan and
visualize what the job looks like. At that time, we cannot go change Hive code
to use TezClient API method to generate these debug artifacts. We will need to
be able to get the same existing code to produce dot plot/DAGplan PB.
bq. one extreme example I can see is that user call submit 2 times in client
side
This depends. Lets say the user call Tezclient.submitDAG() and that error'd out
because of a connection issue. The user code was written to catch any exception
and retry 3 times. In the second try, the user will call TezClient.submitDAG()
on that same DAG. Should this work? It probably should right?
My guess is that we need to consider these cases before we decide on whether
submitDAG() can be called multiple times or not. And what the behavior should
be. IMO, all of this is probably orthogonal to the generate dot plot jira.
Currently, TezClient automatically generates a human readable DAG plan for
debugging purposes when tez.am.log.level is set to true. We should consolidate
that and generation of the dot plot on the client side under a more clear
config like "tez.client.debug.enable" or "tez.client.dry.run" such that these
can be done on any job as and when needed, without needing the user to write
code for it.
> Method verify() of DAG can't been called multiples times in some cases
> ----------------------------------------------------------------------
>
> Key: TEZ-1247
> URL: https://issues.apache.org/jira/browse/TEZ-1247
> Project: Apache Tez
> Issue Type: Bug
> Affects Versions: 0.4.0
> Reporter: Jeff Zhang
> Assignee: Jeff Zhang
> Attachments: Tez-1247-2.patch, Tez-1247.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.2#6252)
