[ 
https://issues.apache.org/jira/browse/HADOOP-13680?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15601540#comment-15601540
 ] 

Steve Loughran commented on HADOOP-13680:
-----------------------------------------

Ooh I'd forgotten about this; I'd one a quick PoC patch and then stashed it 
off, while I was doing HADOOP-13560. I've applied that to branch-2 and stuck it 
up 
[https://github.com/steveloughran/hadoop/tree/s3/HADOOP-13680-get-long-bytes]. 
I don't want to attach the patch or link it as a PR here as it will only cause 
confusion.

Here it is [https://github.com/apache/hadoop/pull/145]

Comparing the two: 
* I added  {{longBytesOption}} to {{S3AUtils}} for broader use; by using it in 
{{getMultipartSizeProperty}}. then {{S3ABlockOutputStream}} gets it for its 
sizes too
* we both patched {{S3AFileSystem}}
* you added a test, which makes it vastly superior.
* neither of us documented this.

How about we combine them: my changes to s3a Test utils, your changes on s3a FS 
and the test, and the docs get updated to indicated that byte values can be 
declared (we can do this in the examples and core-default for completeness); 

...if you do that, I could probably stay untainted enough to be the reviewer....


> fs.s3a.readahead.range to use getLongBytes
> ------------------------------------------
>
>                 Key: HADOOP-13680
>                 URL: https://issues.apache.org/jira/browse/HADOOP-13680
>             Project: Hadoop Common
>          Issue Type: Sub-task
>          Components: fs/s3
>    Affects Versions: 2.8.0
>            Reporter: Steve Loughran
>
> The {{fs.s3a.readahead.range}} value is measured in bytes, but can be 
> hundreds of KB. Easier to use getLongBytes and set to things like "300k"
> This will be backwards compatible with the existing settings if anyone is 
> using them, because the no-prefix default will still be bytes



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

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to