Troubleshooting
Problem
InfoSphere DataStage parallel jobs failed because of "Duplicate name registration" error.
Symptom
Error message in job log.
- Item #: 24
Event ID: 23
Timestamp: 2014-04-18 17:03:34
Type: Fatal
User Name: dsadm
Message Id: IIS-DSEE-TFON-00004
Message: node1: Fatal Error: Duplicate name registration for: diff, APT_DiffOperator.
The error message can show various operator names, to name a few, like APT_TransformOperator, APT_SplitVectOperator, APT_ChangeApplyOperator.
Cause
Duplicate parallel engine libraries exist in the shared library path.
Typically this error happens after restoring Information Server directories using the Copy command to recover from a file system corruption.
There are many shared libraries for the parallel engine in $DSHOME/../PXEngine/lib and $DSHOME/../DSComponents/bin directories. Some libraries are symbolic links to other shared objects in the same directory.
For example, "Duplicate name registration for: diff, APT_DiffOperator" error will happen when
- in AIX, PXEngine/lib/liborchgeneralaix64.so is not a soft link to orchgeneral.o in the same directory.
- in Linux (64-bit), PXEngine/lib/orchgeneral.so is not a soft link to liborchgeneralx86_64.so in the same directory.
Diagnosing The Problem
- Check if there are symbolic links in $DSHOME/../PXEngine/lib and $DSHOME/../DSComponents/bin directory. Normally there should be about 20 symbolic links in each directory.
- Collect and compare the file listing for the PXEngine/lib and DSComponents/bin directories using "ls -l" command from server instances with and without the duplicate name error.
- If you are unable to identify a problem with shared libraries with the above methods, add the environment variable "APT_SHOW_LIBLOAD" as a job parameter to your job and set the value to "1". Then run and review your job log to find any duplicate libraries.
Resolving The Problem
Remove one (or more) of the duplicated shared libraries, for example, by reconstructing the symbolic links.
Note: In MPP configurations, this error can happen after copying the PXEngine directory from an intact node. In order to restore the parallel engine in MPP configuration, using PXEngine/install/copy-orchdist is recommended instead of Copy (cp) command.
Was this topic helpful?
Document Information
Modified date:
16 June 2018
UID
swg21671428