[
https://issues.apache.org/jira/browse/HADOOP-14566?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16059148#comment-16059148
]
Steve Loughran commented on HADOOP-14566:
-----------------------------------------
Note that backward seeks will probably force you to close and re-open the
stream; that's what the object store clients do.
If you do implement seek, it'd mean that readFully() would work, which would
them mean SFTP would be directly usable as source of data for reading things
like HAR files, ORC/Parquet input data, etc. So could be useful. But it'd be
hard work. See also HADOOP-9713.
I really think if you set out to get seek working, especially working fast,
you'll find it hard work. Look at the S3A phase II work of HADOOP-11694 to seem
some of the stuff done there: lazy-seek, programmable forward skip,
fadvise=random, ... . Don't understimate the effort it takes, especially
bearing in mind that a chunk of the testing is done in functional tests
downstream.
> Add seek support for SFTP FileSystem
> ------------------------------------
>
> Key: HADOOP-14566
> URL: https://issues.apache.org/jira/browse/HADOOP-14566
> Project: Hadoop Common
> Issue Type: Improvement
> Components: fs
> Reporter: Azhagu Selvan SP
> Priority: Minor
> Attachments: HADOOP-14566.001.patch, HADOOP-14566.patch
>
>
> This patch adds seek() method implementation for SFTP FileSystem and a unit
> test for the same
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
