From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Maxim Boguk <maxim(dot)boguk(at)gmail(dot)com> |
Cc: | David Rowley <dgrowleyml(at)gmail(dot)com>, pgsql-bugs(at)lists(dot)postgresql(dot)org |
Subject: | Re: BUG #18588: Cannot force/let database use parallel execution in simple case. |
Date: | 2024-08-22 20:42:03 |
Message-ID: | 2073208.1724359323@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-bugs |
I wrote:
> Hmm, you can see both behaviors on the small version of t1, just by
> varying the comparison constant in the WHERE clause. For me, it'll
> use only one worker with "where a<1", and not parallelize at all
> with "where a<0". It looks like it's deciding that it's not worth
> starting workers when too few rows are expected to be returned. That
> would be unsurprising with a normal setting of parallel_setup_cost,
> but it does seem odd with parallel_setup_cost=0.
Ah, I traced through it, and here's what's happening: at small enough
estimated rowcounts, the parallel and non-parallel plans have fuzzily
the same cost (parallel is a shade cheaper, but only a shade).
Their other properties such as pathkeys are the same too. So we get
to the tie-breaking logic in add_path, and what breaks the tie is
the difference in parallel safety: the non-parallel plan is marked
parallel_safe and the parallel one (which by this point is a Gather)
is not.
That tie-break rule is not wrong, because preserving parallel safety
can be a good thing when we come to consider use of the path at higher
plan levels. So I think there's nothing to see here.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Tom Lane | 2024-08-22 20:52:43 | Re: BUG #18588: Cannot force/let database use parallel execution in simple case. |
Previous Message | Maxim Boguk | 2024-08-22 20:27:24 | Re: BUG #18588: Cannot force/let database use parallel execution in simple case. |