JR32317: MODE OVERWRITE DATASETS FAIL TO BE DELETED ON MULITPLE PHYSICAL MACHINES WITH SUBSEQUENT RUN OF DIFFERENT NUMBER OF NODES

APAR status

• Closed as program error.

Error description

• 1. Run PX job twice on a smp machine (smp= one physical machine
  with more than one core) with a two node config file first and
  then a one node config file second. As expected the second run
  works as expected - the dataset from the first run is deleted
  and then created new with two nodes.
  
  2. Run the same job twice more using a configuration file with
  the second node on a different physical machine to the first;
  but as before the first time run use a two node and then use a
  one node configuration file for the second run. This time the
  second run FAILS with the message  ?error when checking
  operator: /tmp/test.ds has 2 partitions, but only 1 is
  accessible from the nodes in the configuration file.?
  

Local fix

Problem summary

• ****************************************************************
  USERS AFFECTED:
  Cluster/grid environments running DataStage parallel jobs
  ****************************************************************
  PROBLEM DESCRIPTION:
  In order to override a dataset, the user needs to make sure
  that all the physical machines where the data files reside must
  be accessible from the parallel configuration file; otherwise,
  an error occurs and the dataset will not be deleted. This causes
  some problems in MPP/cluster/grid environments where data files
  are usually saved in a common file system which is accessible by
  all nodes. This restriction seems unnecessary in those
  environments.
  ****************************************************************
  RECOMMENDATION:
  Patch delivered to the customer
  ****************************************************************
  

Problem conclusion

• The fix is to add an environment variable
  "APT_FORCE_DELETE_DATASET" to lift this restriction. In this
  fix, a nodemap for the delete data file operator is generated
  based on all available nodes in the current configuration file,
  even though not all those physical machines where data files
  reside are present in the current configuration file.
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  WIS DATASTAGE

• Fixed component ID

  5724Q36DS

Applicable component levels

• R753 PSN

     UP