[ 
https://issues.apache.org/jira/browse/CAMEL-7318?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14348482#comment-14348482
 ] 

Grzegorz Grzybek commented on CAMEL-7318:
-----------------------------------------

You're right - for "hdfs://" scheme, the rename just returns {{false}}.

> Concurrency on HDFS Consumer not working efficiently
> ----------------------------------------------------
>
>                 Key: CAMEL-7318
>                 URL: https://issues.apache.org/jira/browse/CAMEL-7318
>             Project: Camel
>          Issue Type: Bug
>          Components: camel-hdfs
>    Affects Versions: 2.11.2
>            Reporter: Martha Obrinteschi
>            Assignee: Grzegorz Grzybek
>            Priority: Minor
>              Labels: concurrency, parallel
>
> If we have two HDFS consumers the files are being processed twice (once by 
> each consumer, waiting one after the other) so the consumers are not working 
> in parallel. If we add this the consumers will work as a team and the 
> transfer will go faster.
> This happens because there is no exception thrown (the rename method just 
> returns true or false and everything carries on as nothing would have 
> happened :).
> In order to fix this we could add in the HdfsInputStream: 49
> boolean tf = info.getFileSystem().rename(new Path(ret.actualPath), new 
> Path(ret.suffixedPath));
> if (!tf) {
>       throw new IOException("This exception is thrown because the rename did 
> not succeded.");
>       }
> And also in the HdfsConsumer: 150
> try {
>      this.rwlock.writeLock().lock();
>      this.istream = 
> HdfsInputStream.createInputStream(fileStatuses[i].getPath().toString(), 
> this.config);
> } catch (IOException ioe) {
>      log.info(ioe.getMessage()        + " If the rename fails we move on to 
> the next file.");
>      continue;
> }
>   



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to