Scheduling Batch Unix And Linux Production Jobs; Improving Performance In A Unix Production Environment; Performing Incremental Updates On Unix And Linux; Example - Updating A Model Using An Mdl Script - IBM Cognos User Manual

v You can view the contents of the log file, which contains messages issued by
Cognos Transformer.
v If you used a crontab file or scheduled the job with the UNIX or Linux at
command, you can check your email for a completion message, and review the
log file for any warnings or errors.
v You can check your ModelSaveDirectory location for a checkpoint file (.qy?).
Because Cognos Transformer automatically deletes checkpoint files when
processing ends successfully, a checkpoint file means that a suspended model
exists. Check the log file for errors associated with the processing of that model.
v If you created a shell script that includes cogtr commands, you can check the
exit status to detect operations that did not end successfully. An exit status value
of 0 indicates successful completion. Any other value indicates an error.
By default, all messages generated by Cognos Transformer are directed to the
standard output stream. You can direct them to a log file instead, by setting the
LogFileName preference in the cogtr.xml file:
<Preference Name="LogFileName" Value="file_name"/>
You can analyze the error and warning messages in the log file to help you isolate
problems encountered when Cognos Transformer attempted to read the data
source or write data to the cube. For more information, see "Log Files" on page
196.

Scheduling Batch UNIX and Linux Production Jobs

If you want to automate cube creation, you can use the UNIX or Linux scheduling
commands, the UNIX/Linux at command, or a crontab file, to schedule these jobs
at convenient times, in batch mode.
For example, suppose you have a model that you use to create 20 individual cubes
and a cube group consisting of 10 cubes. You can supply scheduling information to
enable and disable the creation of specific cubes at specific times.

Improving Performance in a UNIX Production Environment

Because .pyj files load faster than .mdl files, consider converting your .mdl files to
.pyj files in a UNIX production environment.
To complete the conversion, use the following command line syntax:
cogtr -m model.mdl -s model.pyj
After creating the .pyj files, create cubes using the following command line syntax:
cogtr -c -p model.pyj

Performing Incremental Updates on UNIX and Linux

If you update cubes incrementally and a cube update fails, if you try to repeat the
most recent incremental update, some records from the update may already have
been written to the cube. Restarting the process causes Cognos Transformer to add
these records to the cube twice, producing inaccuracies.

Example - Updating a Model Using an MDL Script

This example demonstrates how to use an MDL script to split the generation of
categories and the creation of cubes into separate processes.
Chapter 9. Guidelines for Optimizing Production Environments
201