[jira] [Commented] (CRUNCH-658) Add a way to skip the getSize checks for Sources from object stores

["Andrew Olson (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CRUNCH-658?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16772200#comment-16772200 ] 

Andrew Olson commented on CRUNCH-658:
-------------------------------------

Looks like skipping the getLastModifiedAt for sources & targets is another potential optimization? Especially since it's only actually used (as far as I can tell) if the WriteMode is CHECKPOINT. The implementations of getPathSize and getLastModifiedAt in SourceTargetHelper are similar so I would expect them to have the same performance issue with object stores.

> Add a way to skip the getSize checks for Sources from object stores
> -------------------------------------------------------------------
>
>                 Key: CRUNCH-658
>                 URL: https://issues.apache.org/jira/browse/CRUNCH-658
>             Project: Crunch
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 0.14.0
>            Reporter: Josh Wills
>            Assignee: Josh Wills
>            Priority: Major
>
> Ran into a problem when using Crunch to process a _lot_ of data from S3: the getSize checks can be very slow to run and don't materially add much to the overall processing of a pipeline when things like reducer counts are manually specified. I'd like to add a way to disable the file size checks, either globally or for specific input sources.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)