16.20 - Additional Debugging Strategies for Complex Job Failures - Parallel Transporter

Teradata® Parallel Transporter User Guide

Product
Parallel Transporter
Release Number
16.20
Published
August 2020
Language
English (United States)
Last Update
2020-08-27
dita:mapPath
uah1527114222342.ditamap
dita:ditavalPath
Audience_PDF_product_tpt_userguide_include.ditaval

In some cases, simply evaluating the job logs and error tables does not provide enough information to adequately define the required corrective action. In other cases, the corrective action is in place, but the job still does not run correctly. In these cases, Teradata PT provides two additional levels of debugging:

  • Check the values of system resources such as shared memory, processes, semaphores, memory, and so on.

    For example, on Solaris running on a SPARC system, use the following commands to get the values:

    • /usr/sbin/sysdef -i | grep SHMMAX
    • /usr/sbin/sysdef -i | grep SHMSEG
    • /usr/sbin/sysdef -i | grep SEMMNI
    • /usr/sbin/sysdef -i | grep SEMMNS
    • ulimit -a
  • Run the job in trace mode
    • tbuild -t -f <filename>
    • Run the operators in your Teradata PT job in trace mode using the TraceLevel attribute.
      TraceLevel = ‘all’
  • Provide truss output (UNIX system only) from the Teradata PT problem component if any of the following errors occurs:
    • IPC Initialization Error (Inter-Process Communication problem)
    • Failed to create Coordinator task
    • Unexpected hanging
  • Use the following steps to get the truss output of the problem component:
    1. ps -ef | grep tbuild (if Coordinator, or Executor).
    2. Find the processid for the problem component.
    3. truss -f -o /tmp/trussout -p <processid>.