Re: pb with join plan

On 6/21/23 00:26, Marc Millas wrote:
>
>
> On Tue, Jun 20, 2023 at 11:19 PM David Rowley <dgrowleyml(at)gmail(dot)com
> <mailto:dgrowleyml(at)gmail(dot)com>> wrote:
>
> On Wed, 21 Jun 2023 at 08:34, Marc Millas <marc(dot)millas(at)mokadb(dot)com
> <mailto:marc(dot)millas(at)mokadb(dot)com>> wrote:
> >
> > On Tue, Jun 20, 2023 at 10:14 PM David Rowley
> <dgrowleyml(at)gmail(dot)com <mailto:dgrowleyml(at)gmail(dot)com>> wrote:
> >>
> >> On Wed, 21 Jun 2023 at 07:42, Marc Millas <marc(dot)millas(at)mokadb(dot)com
> <mailto:marc(dot)millas(at)mokadb(dot)com>> wrote:
> >> > But if I do the same with clause one OR clause 2, I have to 
> kill the request after an hour, seeing the filesystem showing more
> than 140 Mb of increased usage.
> >>
> >>
> > link to the anonymized plan of the req with one clause :
> https://explain.depesz.com/s/TWp4 <https://explain.depesz.com/s/TWp4>
>
> link to the plan with the second
> clause alone: https://explain.depesz.com/s/byW5
> <https://explain.depesz.com/s/byW5> 
> link to the plan with both clauses ORed (the one not
> finishing) https://explain.depesz.com/s/jHO2
> <https://explain.depesz.com/s/jHO2>
>
>
>
> It's quite difficult to know what the problem is you want to fix here.
> Your initial post indicated it was the query with the OR condition
> that was causing you the problems, but the plan you've posted has no
> OR condition?!
>
> You're more likely to get help here if you take time to properly
> explain the situation and post the information that's actually
> relevant to the problem you're having, or state the problem more
> clearly, as there's a mismatch somewhere.
>
> It might also be worth having a look at
> https://wiki.postgresql.org/wiki/Slow_Query_Questions
> <https://wiki.postgresql.org/wiki/Slow_Query_Questions> . EXPLAIN is not
> going to tell us what part of the query is slow. I'll let the wiki
> page guide you into what to do instead.
>
>  
> I know that page. obviously, as I have to kill the request, I cannot
> provide a explain analyze... 
>

It's a bit weird the "victor" table is joined seemingly without any join
conditions, leading to a cross join (which massively inflates the cost
for joins above it). Maybe the anonymized plan mangles it somehow.

regards

--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company