[
https://issues.apache.org/jira/browse/FLINK-1006?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14076658#comment-14076658
]
ASF GitHub Bot commented on FLINK-1006:
---------------------------------------
Github user warneke commented on the pull request:
https://github.com/apache/incubator-flink/pull/71#issuecomment-50386469
When you build Flink with the hadoop2 version set to 2.3.0-mapr-4.0.0-FCS,
all dependencies are included in the library folder automatically through the
regular Maven dependency resolution. There are no special configuration
settings necessary. To be honest, I never tried to use the vanilla build and
add the dependencies for MapR later on.
If people don't want to compile Flink themselves, I assume they are also
not too keen on manually fixing Hadoop dependencies. How about we provide
precompiled versions of Flink releases for the three major Hadoop players? I
think you guys started it for Cloudera anyway, right? I would volunteer to
cover the MapR part.
> Add support for MapR file system
> --------------------------------
>
> Key: FLINK-1006
> URL: https://issues.apache.org/jira/browse/FLINK-1006
> Project: Flink
> Issue Type: Improvement
> Components: Distributed Runtime
> Affects Versions: 0.6-incubating
> Reporter: Daniel Warneke
> Assignee: Daniel Warneke
> Priority: Minor
>
> I would like to add support for the MapR file system to Flink. The MapR file
> system is a propriatory distributed file system which is interface-compatible
> to HDFS and core to the MapR Hadoop distribution.
> Support for MapR FS would require one additional class
> eu.stratosphere.runtime.fs.maprfs.MapRFileSystem which basically tells Flink
> how to deal with the URL scheme name "maprfs://" and connect to the container
> databases. For everything else, we could reuse the already existing HDFS
> wrappers.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
