hadoop-common-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Omkar Joshi <ojo...@hortonworks.com>
Subject Re: New Distributed Cache
Date Thu, 11 Jul 2013 18:08:57 GMT
Yeah Andrew.. there seems to be some problem with context.getCacheFiles()
api which is returning null..

 Path[] cachedFilePaths =

          context.getLocalCacheFiles(); // I am checking why it is
deprecated...

      for (Path cachedFilePath : cachedFilePaths) {

        File cachedFile = new File(cachedFilePath.toUri().getRawPath());

        System.out.println("cached fie path >> "

            + cachedFile.getAbsolutePath());

      }

I hope this helps for the time being.. JobContext was suppose to replace
DistributedCache api (it will be deprecated) however there is some problem
with that or I am missing something... Will reply if I find the solution to
it.

context.getCacheFiles will give you the uri used for localizing files...
(original uri used for adding it to cache)... However you can use
DistributedCache.getCacheFiles() api till context api is fixed.

context.getLocalCacheFiles .. will give you the actual file path on node
manager... (after file is localized).

Thanks,
Omkar Joshi
*Hortonworks Inc.* <http://www.hortonworks.com>


On Thu, Jul 11, 2013 at 8:19 AM, Botelho, Andrew <Andrew.Botelho@emc.com>wrote:

> So in my driver code, I try to store the file in the cache with this line
> of code:****
>
> ** **
>
> job.addCacheFile(new URI("file location"));****
>
> ** **
>
> Then in my Mapper code, I do this to try and access the cached file:****
>
> ** **
>
> URI[] localPaths = context.getCacheFiles();****
>
> File f = new File(localPaths[0]);****
>
> ** **
>
> However, I get a NullPointerException when I do that in the Mapper code.**
> **
>
> ** **
>
> Any suggesstions?****
>
> ** **
>
> Andrew****
>
> ** **
>
> *From:* Shahab Yunus [mailto:shahab.yunus@gmail.com]
> *Sent:* Wednesday, July 10, 2013 9:43 PM
> *To:* user@hadoop.apache.org
> *Subject:* Re: New Distributed Cache****
>
> ** **
>
> Also, once you have the array of URIs after calling getCacheFiles  you
> can iterate over them using File class or Path (
> http://hadoop.apache.org/docs/current/api/org/apache/hadoop/fs/Path.html#Path(java.net.URI)
> )****
>
> ** **
>
> Regards,****
>
> Shahab****
>
> ** **
>
> On Wed, Jul 10, 2013 at 5:08 PM, Omkar Joshi <ojoshi@hortonworks.com>
> wrote:****
>
> did you try JobContext.getCacheFiles() ?****
>
> ** **
>
>
> ****
>
> Thanks,****
>
> Omkar Joshi****
>
> *Hortonworks Inc.* <http://www.hortonworks.com>****
>
> ** **
>
> On Wed, Jul 10, 2013 at 10:15 AM, Botelho, Andrew <Andrew.Botelho@emc.com>
> wrote:****
>
> Hi,****
>
>  ****
>
> I am trying to store a file in the Distributed Cache during my Hadoop job.
> ****
>
> In the driver class, I tell the job to store the file in the cache with
> this code:****
>
>  ****
>
> Job job = Job.getInstance();****
>
> job.addCacheFile(new URI("file name"));****
>
>  ****
>
> That all compiles fine.  In the Mapper code, I try accessing the cached
> file with this method:****
>
>  ****
>
> Path[] localPaths = context.getLocalCacheFiles();****
>
>  ****
>
> However, I am getting warnings that this method is deprecated.****
>
> Does anyone know the newest way to access cached files in the Mapper code?
> (I am using Hadoop 2.0.5)****
>
>  ****
>
> Thanks in advance,****
>
>  ****
>
> Andrew****
>
> ** **
>
> ** **
>

Mime
View raw message