[
https://issues.apache.org/jira/browse/PIG-4060?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Cheolsoo Park updated PIG-4060:
-------------------------------
Description:
Currently, TezJob extends ControlledJob, and TezLauncher submits TezJob to the
cluster via ControlJob. However, there is no reason for this since ControlJob
always contains a single job in Tez unlike MR.
There are some operators such as MergeJoin that consists of more than 1
disjoint DAGs. But this case is handled by TezPlanContainer. TezPlanContainer
holds multiple DAGs, and TezLauncher compiles and submits them one by one in
order.
I propose we should rewrite TezJob and TezLauncher using ExecutorService.
was:
Currently, TezJob extends ControlledJob, and TezLauncher submits TezJob to the
cluster via ControlJob. However, there is no reason for this since ControlJob
always contains a single job in Tez unlike MR.
There are some operators such as MergeJoin that consists of more than 1
disjoint DAGs. But this case is handled by TezPlanContainer. That is,
TezPlanContainer keeps multiple DAGs, and TezLauncher compiles/submits them one
by one in order.
> Refactor TezJob and TezLauncher
> -------------------------------
>
> Key: PIG-4060
> URL: https://issues.apache.org/jira/browse/PIG-4060
> Project: Pig
> Issue Type: Sub-task
> Components: tez
> Reporter: Cheolsoo Park
> Assignee: Cheolsoo Park
> Fix For: 0.14.0
>
>
> Currently, TezJob extends ControlledJob, and TezLauncher submits TezJob to
> the cluster via ControlJob. However, there is no reason for this since
> ControlJob always contains a single job in Tez unlike MR.
> There are some operators such as MergeJoin that consists of more than 1
> disjoint DAGs. But this case is handled by TezPlanContainer. TezPlanContainer
> holds multiple DAGs, and TezLauncher compiles and submits them one by one in
> order.
> I propose we should rewrite TezJob and TezLauncher using ExecutorService.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
