Hi Silvan,

Are you trying to run MapReduce? You could look at 
createApplicationSubmissionContext and createApplicationResource in 
YarnRunner.java. Have you overridden the resolvePath function in your 
filesystem implementation?

-Varun

From:  Silvan Kaiser
Reply-To:  "[email protected]<mailto:[email protected]>"
Date:  Friday, May 8, 2015 at 3:52 PM
To:  "[email protected]<mailto:[email protected]>"
Subject:  URI missing scheme and authority in job start with new FileSystem 
implementation


Hello!
I am running a new FileSystem implementation, testing it with a simple 
standalone operation from the Hadoop "Single Node Setup" page. Staging example 
jobs works alright but upon scheduling the container management fails because 
of an URI without scheme
(and authority).

NodeManager log snippet:
...

2015-05-08 15:18:15,331 INFO 
org.apache.hadoop.yarn.server.nodemanager.containermanager.container.ContainerImpl:
 Got exception parsing jobSubmitDir/job.splitmetainfo and value resource { 
port: -1 file: 
"/tmp/hadoop-yarn/staging/kaisers/.staging/job_1430315033077_0086/job.splitmetainfo"
} size: 771 timestamp: 1431091092000 type: FILE visibility: APPLICATION
2015-05-08 15:18:15,331 WARN 
org.apache.hadoop.yarn.server.nodemanager.containermanager.container.ContainerImpl:
 Failed to parse resource-request
java.net.URISyntaxException: Expected scheme name at index 0: 
:///tmp/hadoop-yarn/staging/kaisers/.staging/job_1430315033077_0086/job.splitmetainfo

...


The path element of the URI is correct, it simply is missing it's scheme and 
authority but i cannot find out why so far.
Please note that the URIs used during staging are correct and contain scheme & 
authority correctly. This means the incomplete URI is created during job 
creation.

My guess so far: this is either because I am missing some configuration for the 
Nodemanager (therefore he does not know about my FileSystem implementation and 
throws the unknown scheme & authority out the window) or that the FileSystem 
implementation has
an error / did not override a method that requires adoption.

Can someone please give me a hint where the URIs used by the NodeManager are 
constructed prior to scheduling? This would allow me to trace back where the 
incomplete URI is coming from...

Best regards
Silvan


--
Quobyte GmbH
Boyenstr. 41 - 10115 Berlin-Mitte - Germany
+49-30-814 591 800 - www.quobyte.com 
<http://www.quobyte.com/><http://www.quobyte.com/>
Amtsgericht Berlin-Charlottenburg, HRB 149012B
Management board: Dr. Felix Hupfeld, Dr. Björn Kolbeck, Dr. Jan Stender









--
Quobyte GmbH
Hardenbergplatz 2 - 10623 Berlin - Germany
+49-30-814 591 800 - www.quobyte.com <http://www.quobyte.com/>Amtsgericht 
Berlin-Charlottenburg, HRB 149012B
management board: Dr. Felix Hupfeld, Dr. Björn Kolbeck, Dr. Jan Stender

Reply via email to