[
https://issues.apache.org/jira/browse/HDFS-15377?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17119218#comment-17119218
]
Yang Yun commented on HDFS-15377:
---------------------------------
Thanks [~elgoiri] for the review.
Updated to HDFS-15377.003.patch with following changes,
* Add private method createBlockSender for the new code.
* declare the variable 'curPosition' as AtomicLong for find bug issue.
> BlockScanner scans one part per round, expect full scans after several rounds
> -----------------------------------------------------------------------------
>
> Key: HDFS-15377
> URL: https://issues.apache.org/jira/browse/HDFS-15377
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: Yang Yun
> Assignee: Yang Yun
> Priority: Minor
> Attachments: HDFS-15377.002.patch, HDFS-15377.003.patch
>
>
> For reducing disk IO, one block is separated to multiple parts, BlockScanner
> scans only one part per round. Expect that after several rounds, the full
> block should be scanned
> Add a new option "dfs.block.scanner.part.size". the maximum data size per
> scan by the block scanner. this value should be the multiple of chunk size,
> for example, 512, 1024, 4096 ...
> Default value is -1, will disable partial scan.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
