[ 
https://issues.apache.org/jira/browse/HDFS-1900?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13135301#comment-13135301
 ] 

Konstantin Shvachko commented on HDFS-1900:
-------------------------------------------

I think "dfs.block.size" introduced in HADOOP-4952 was just a mistake. I think 
the intention was to have it "dfs.blocksize", same as in HDFS. I see a bunch of 
other file system keys in FsConfig, that duplicate their HDFS counterparts, but 
the naming is consistent across them. So I assume that extra dot between block 
and size was a typo.
Having said that I think it is not a good idea to duplicate key definitions. 
The reason is exactly the typos or inconsistent renaming of those properties or 
the default values. E.g. in common 
{{FsConfig.FS_DEFAULT_BLOCK_SIZE = 32 MB}} 
while in HDFS 
{{DFSConfigKeys.DFS_BLOCK_SIZE_DEFAULT = 64 MB}}
This is really messy.

The only method from {{FsConfig}} that is used in the code is 
{{getDefaultFsURI()}}. I propose to remove everything else from {{FsConfig}} in 
order to avoid confusion. 
The rational behind this is that {{FsConfig}} should only contain keys that are 
specified in core-site.xml. The keys that belolng to hdf-site.xml shoud be 
described in {{DFSConfigKeys}}.
It also looks that Tom's documentation change HDFS-671 describes keys 
consistently with this assumption.
                
> Use the block size key defined by common 
> -----------------------------------------
>
>                 Key: HDFS-1900
>                 URL: https://issues.apache.org/jira/browse/HDFS-1900
>             Project: Hadoop HDFS
>          Issue Type: Improvement
>    Affects Versions: 0.21.1
>            Reporter: Eli Collins
>            Assignee: Abel Perez
>              Labels: newbie
>             Fix For: 0.22.0
>
>         Attachments: HDFS-1900.txt
>
>
> HADOOP-4952 added a dfs.block.size key to common configuration, defined in 
> o.a.h.fs.FsConfig. This conflicts with the original HDFS block size key of 
> the same name, which is now deprecated in favor of dfs.blocksize. It doesn't 
> make sense to have two different keys for the block size (ie they can 
> disagree). Why doesn't HDFS just use the key defined in common?

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to