Phantom Error - Caught unknown exception

[thompsonp]

We have a number of jobs that intermittently abort with the following errors.

main_program: Fatal Error: Caught unknown exception in main program: terminating.
Contents of phantom output file =>
DataStage Job 698 Phantom 8676

Looking in the &PH& directory we find a couple of files with the right timestamp, contents shown below:

File 1

DataStage Job 698 Phantom 8676
readSocket() returned 16
DataStage Phantom Finished.


File2

DataStage Job 698 Phantom 6960
Program "DSD.RUN": Line 2098, Variable previously undefined. Zero length string used.
[same error repeated several times then]
Program "DSD.RUN": Line 2098, Variable previously undefined. Zero length string used.
The variable "APT_MSG_FILELINE" is not in the environment.
The variable "APT_RECORD_COUNTS" is not in the environment.
The variable "APT_SHOW_COMPONENT_CALLS" is not in the environment.
The variable "APT_STARTUP_STATUS" is not in the environment.
The variable "APT_MSG_FILELINE" is not in the environment.
The variable "APT_RECORD_COUNTS" is not in the environment.
The variable "APT_SHOW_COMPONENT_CALLS" is not in the environment.
The variable "APT_STARTUP_STATUS" is not in the environment.

15 record(s) selected to SELECT list #1.
Program "DSD.OshRun": Line 1328, Unable to unlink the operating system file "DSD.OshMonitor_57523_14143".
DataStage Phantom Finished.

Can anyone advise on what the cause of this could be or how to track down the cause.
The one thing in common with the jobs that report this problem is the use of a shared container (but not all jobs using the container produce this error).

Removing the container, or making it local resolves the problem (as far as we can tell given that it is intermittent with the shared container in the job).

Thanks
Paul

[mabramson]

There are two reasons why you might be getting those errors. 1) Someone changed the environment settings in Datastage Administrator. Some of those are default environment variables that are set up for reporting in PX. If you changed the value from "False" to "True", then when the PX jobs are executed, it will look to fill these values. If it can't find them you will get this error. 2) Another reason for this error, if the first reason doesn't apply, is because the user who is running these jobs does not have the proper permissions. Check an existing user's permissions for Datastage and compare them.

Try those items. That should correct your problem.

MA

[thompsonp]

Some of the reporting environment variables were changed from False to True in the environment the jobs have been moved to. However I don't know why that would cause some jobs to have intermittent problems whilst others work as before.

How would it be possible to change these values in Administrator and then for DataStage not to find them when the job runs? Is it possible that setting some environment variables then looks for others to be set as well that may not be in our DSParams file?

There is also a new user running all the jobs, so differences in setup and permissions are possible. Again the intermittent nature of the problem led me to believe this was unlikely to be the cause.

I'll have to investigate further when I'm back at work tomorrow.

[radar_etl]

Was there a solution to this problem?