GPFS Connection Error - NoInputPathFoundExeception

Symptoms

When trying to access data on GPFS, the following warning and error is observed in the conductor.log file:

WARN [2014-05-02 09:55:20] (Summary.java:46) - Can not parse input sample: InputFormat: org.apache.hadoop.mapred.TextInputFormat,OutputFormat: org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat,Location: gpfs:/path/to/my/hdfs/data

...

ERROR [2014-05-02 10:10:28] (DefineFieldsController.java:216) - Can not parse input sample

 datameer.dap.sdk.importjob.NoInputPathFoundException: Could not find any existing file for ...

 

The data that is being attempted to be accessed is owned by a user other than the user that runs Datameer. The user that runs Datameer is a member of the group 

Cause


The libgpfsbiginsights.so library file contained in the native/lib/Linux-amd64-64 directory is suspected to have a bug related to group permissions on GPFS.

Resolution

Replace the libgpfsbiginsights.so library file with a newer version that does not contain the bug. Please consult with IBM Support for further information on the latest compatible library version.


For reference, the following library versions were noted:

Properly Accesses Group PermissionsFilenamemd5sum
No libgpfsbiginsights.so 5bbc10baf5061c31a75f444e7451bd0d
Yes libgpfsbiginsights.so c9abe71019eec513f0f861e93bee660f