[
https://issues.apache.org/jira/browse/PIG-3539?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Daniel Dai updated PIG-3539:
----------------------------
Attachment: PIG-3539-1.patch
The patch includes:
1. TezPlanContainer & associated visitors
2. TezOperPlan will be segmented into TezPlanContainer. The segment point is
TezOperator.segmentBelow
3. TezSessionManager implement a session pool
4. TezResourceManager manage the localResources for session/DAG
5. Asynchonous DAG submission/monitoring
6. Upgrade commons-cli since it has conflict with tez (Otherwise
TestTezLauncher fail due to invalid command line option)
There is no tests for the new components right now, I will add some later.
Existing tests all pass.
> Pig should be able to submit multiple DAG
> -----------------------------------------
>
> Key: PIG-3539
> URL: https://issues.apache.org/jira/browse/PIG-3539
> Project: Pig
> Issue Type: Sub-task
> Components: tez
> Reporter: Daniel Dai
> Assignee: Daniel Dai
> Fix For: tez-branch
>
> Attachments: PIG-3539-1.patch
>
>
> There could be several scenario Pig want to submit multiple DAG:
> 1. a blocking operator in grunt
> 2. a blocking operator in script (exec, fs)
> 3. break a DAG into two for possible inner DAG optimization
> Pig should use TezSession so that we can reuse AM. TezSession should be a
> thread local variable. For those who use Pig Java API, they will use the same
> AM within a thread, but different AM in another thread.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
