Thanks for the feedback.
Posting an updated set of patches. Changes are based on feedback, as
detailed below:
There's a couple of potential issues currently being looked at:
- locking issues in additional parallel-safety checks?
- apparent uneven work distribution across the parallel workers, for
large insert data
[Antonin]
- Fixed bad Assert in PrepareParallelMode()
- Added missing comment to explain use of GetCurrentCommandId() in
PrepareParallelMode()
- Some variable name shortening in a few places
- Created common function for creation of non-parallel and parallel
ModifyTable paths
- Path count variable changed to bool
- Added FIXME comment to dubious code for creating Gather target-list
from ModifyTable subplan
- Fixed check on returningLists to use NIL instead of NULL
[Amit]
- Moved additional parallel-safety checks (for modify case) into
max_parallel_hazard()
- Removed redundant calls to max_parallel_hazard_test()
- Added Asserts to "should never happen" null-attribute cases (and
added WARNING log missing from one case)
- Added comment for use of NoLock in max_parallel_hazard_for_modify()
[Vignesh]
- Fixed a couple of typos
- Added a couple of test cases for testing that the same transaction
is used by all parallel workers
Regards,
Greg Nancarrow
Fujitsu Australia