[ https://issues.apache.org/jira/browse/HADOOP-12444?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Steve Loughran updated HADOOP-12444:
------------------------------------
    Attachment: HADOOP-12444-007.patch

This is patch 007, which is patch 006 applied to the latest (006) patch of the positioned readable tests.

h3. in {{read(byte[] buf, int off, int len)}}

* when reading a len==0 from a zero byte file at offset 0, then return value must be zero
* the position update code had been pulled into the try/catch clause; the pos counters would have become invalid if an exception was raised and the re-open and read() worked

h3. other

# use multiple catch to catch the socket and socket timeout exceptions; simpler codepath.
# in the refactored {{closeStream()}} operation, implemented HADOOP-11874: if  an IOE is thrown in {{wrappedStream.close()}} it's caught, the operation converted to an abort().

As well as all the tests passing, I spent time reviewing the exception handling code, to see if a problem was lurking which tests weren't catching. Remember: object stores are often used long-haul, so do fail —recoverability matters.

I'm happy with the code as it is now: LGTM

> Consider implementing lazy seek in S3AInputStream
> -------------------------------------------------
>
>                 Key: HADOOP-12444
>                 URL: https://issues.apache.org/jira/browse/HADOOP-12444
>             Project: Hadoop Common
>          Issue Type: Sub-task
>          Components: fs/s3
>    Affects Versions: 2.7.1
>            Reporter: Rajesh Balamohan
>            Assignee: Rajesh Balamohan
>         Attachments: HADOOP-12444-004.patch, HADOOP-12444-005.patch, HADOOP-12444-006.patch, HADOOP-12444-007.patch, HADOOP-12444.1.patch, HADOOP-12444.2.patch, HADOOP-12444.3.patch, HADOOP-12444.WIP.patch, hadoop-aws-test-reports.tar.gz
>
>
> - Currently, "read(long position, byte[] buffer, int offset, int length)" is not implemented in S3AInputStream (unlike DFSInputStream). So, "readFully(long position, byte[] buffer, int offset, int length)" in S3AInputStream goes through the default implementation of seek(), read(), seek() in FSInputStream. 
> - However, seek() in S3AInputStream involves re-opening of connection to S3 everytime (https://github.com/apache/hadoop/blob/trunk/hadoop-tools/hadoop-aws/src/main/java/org/apache/hadoop/fs/s3a/S3AInputStream.java#L115).  
> - It would be good to consider having a lazy seek implementation to reduce connection overheads to S3. (e.g Presto implements lazy seek. https://github.com/facebook/presto/blob/master/presto-hive/src/main/java/com/facebook/presto/hive/PrestoS3FileSystem.java#L623)



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