Changes to TM1 and batch update model in 9.4
Issue: In TM1 9.4 the locking and batch update models have been comprehensively overhauled. During a TI chore or process all data changes are held in a temporary memory space and committed in a single batch upon successful completion of the chore or process. Any log file entries are also held in temporary memory and only committed to the tm1s.log file when the cubes are batch updated. If a major error is encountered during a TI process then no data changes are committed and no log entries are written. This is part of the process known as the "rollback" mechanism. Should cube logging be left on during a load process, especially for a lengthy process or cumulative load the temporary memory consumed by the virtual log file can become excessively large and cause a "system out of memory" server issue.
In prior versions of TM1 this would not have happened as log entries were written immediately and not upon successful completion of the cube batch update.
The solution to this problem is simple, ensure that cube logging is switched off for all batch update processes.