[
https://issues.apache.org/jira/browse/HDFS-6898?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14109966#comment-14109966
]
Colin Patrick McCabe commented on HDFS-6898:
--------------------------------------------
bq. Colin Patrick McCabe Will fallocate trigger disk I/O? I thought it only
manipulates metadata: http://linux.die.net/man/1/fallocate – "this is done
quickly by allocating blocks and marking them as uninitialized, requiring no IO
to the data blocks"
It depends on the underlying file system. For a filesystem which uses extents,
like ext4, not much I/O would be required. For something like ext2, you have
to allocate each ext2 block individually and link it to the rest of them, which
could be quite a lot of I/O. That's a fair point, though... it will not be 128
MB of I/O in either case... it would be less than that. It's all metadata I/O
essentially.
bq. However we still have the other issue that Colin mentioned - we cannot
deduce the end of file on restart since our block file format lacks any
header/meta information.
Yeah.
bq. Colin Patrick McCabe, the attached .03.patch is current.
Thanks. Will try to take a look later today
> DN must reserve space for a full block when an RBW block is created
> -------------------------------------------------------------------
>
> Key: HDFS-6898
> URL: https://issues.apache.org/jira/browse/HDFS-6898
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode
> Affects Versions: 2.5.0
> Reporter: Gopal V
> Assignee: Arpit Agarwal
> Attachments: HDFS-6898.01.patch, HDFS-6898.03.patch
>
>
> DN will successfully create two RBW blocks on the same volume even if the
> free space is sufficient for just one full block.
> One or both block writers may subsequently get a DiskOutOfSpace exception.
> This can be avoided by allocating space up front.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
