You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@tez.apache.org by "Rajesh Balamohan (JIRA)" <ji...@apache.org> on 2015/04/10 07:47:12 UTC

[jira] [Updated] (TEZ-1752) Inputs / Outputs in the Runtime library should be interruptable

     [ https://issues.apache.org/jira/browse/TEZ-1752?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Rajesh Balamohan updated TEZ-1752:
----------------------------------
    Attachment: TEZ-1752.1.patch

[~sseth] - Please review when you have time. 
- Tested it on cluster with fair number of jobs and ensured that resources are not leaked.
- interrupts are properly propagated to higher level for proper cleanup
- This still doesn't handle certain scenarios when fetchers are blocked on connections (server busy, blocked on IO etc) which needs to be addressed as a part of TEZ-711.
- For merger, interrupt state is checked before reporting the progress in TezMerger.writeFile().  This should give headroom to check if its interrupted in the middle of merging. 
- For unordered inputs, stale file cleaning (DiskFetchedInput, LocalDiskFetchedInputs) are not cleaned when interrupted.  However, these would be automatically be cleaned up as part of job cleanup.

> Inputs / Outputs in the Runtime library should be interruptable
> ---------------------------------------------------------------
>
>                 Key: TEZ-1752
>                 URL: https://issues.apache.org/jira/browse/TEZ-1752
>             Project: Apache Tez
>          Issue Type: Improvement
>            Reporter: Siddharth Seth
>         Attachments: TEZ-1752.1.patch
>
>
> Not possible to preempt tasks without killing containers without this.
> There's still the problem of Processors not supporting interrupts. We may need API enhancements to either query IPOs on whether they are interrupbtible.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)