Horton

Re: unstable query plan on pg 16,17,18

Laurenz Albe <laurenz.albe@cybertec.at>

View thread View on postgr.es
From: Laurenz Albe <laurenz.albe@cybertec.at>
To: Attila Soki <atiware@gmx.net>
Cc: pgsql-performance@postgresql.org
Date: 2026-02-23T21:44:00Z
Lists: pgsql-performance 
On Mon, 2026-02-23 at 22:19 +0100, Attila Soki wrote:
> To "switch" between the ok and wrong plans I execute
> vacuumdb -Upostgres -vZ -t schema1.table_k db1
> multiple times (1x-3x) until the plan flips. So I thought the table should already have been analyzed.
> In my first email I called table_k as the "tbl_used_in_query"

Ok, I see.  Then a simple ANALYZE won't do.

Perhaps the column needs more detailed statistics?

  ALTER TABLE schema1.table_k ALTER dp_end_dat SET STATISTICS 1000;
  ANALYZE schema1.table_k (dp_end_dat);

> or have I misunderstood that with the analyze?

No; I missed that point.

Yours,
Laurenz Albe