On Wednesday, November 17, 2021 4:19 PM, Guillaume Lelarge guillaume@lelarge.info wrote:
>> -> Parallel Seq Scan on c (cost=0.00..8591.67 rows=416667 width=0) (actual time=0.030..140.036 rows=333333 loops=3)
>In my previous example, actual row number is 333333*3=1e6(which is correct), so I think the actual time is 140.036*3ms.
>Do your think the loops(3) has no meaning for parallel scan node when calculate actual time?
>
>As far as I understand it, you have to multiply the number of rows by the number of loops, but this doesn't apply to duration at least for parallel >queries.
Yes, I got your point.
I’m not familiar with PostgreSQL planner/executor, but if the code is correct, then maybe some modification should be done at [1] as below:
Before:
Multiply by the loops value to get the total time actually spent in the node.
After:
Multiply by the loops value to get the total time actually spent in the node (for node in the parallel portion of the plan, this is not needed).
[1] https://www.postgresql.org/docs/14/using-explain.html
What do you think?
Regards,
Tang