Re: Partial index plan/cardinality costing

[James Coleman <jtc331@xxxxxxxxx>]

Bump, and curious if anyone on hackers has any ideas here: of particular interest is why the (pk, created_at) index can possibly be more valuable than the (created_at, pk) variant since the former effectively implies having to scan the entire index.

On Fri, Sep 7, 2018 at 12:17 PM James Coleman <jtc331@xxxxxxxxx> wrote:

I have the following tables:

- m(pk bigserial primary key, status text): with a single row

- s(pk bigserial primary key, status text, action_at date, m_fk bigint):

  * 80% of the data has action_at between the current date and 1 year ago

     and status of E or C

  * 20% of the data has action_at between 5 days ago and 25 days into the

     future and status of P, PD, or A

I have two partial indexes:

- s_pk_action_at on s(pk, action_at) where status in ('P', 'PD', 'A')

- s_action_at_pk on s(action_at, pk) where status in ('P', 'PD', 'A')

With the query:

SELECT s.pk FROM s

INNER JOIN m ON m.pk = s.m_fk

WHERE

  s.status IN ('A', 'PD', 'P')

  AND (action_at <= '2018-09-06')

  AND s.status IN ('A', 'P')

  AND m.status = 'A';

I generally expect the index s_action_at_pk to always be preferred over s_pk_action_at. And on stock Postgres it does in fact use that index (with a bitmap index scan).

We like to set random_page_cost = 2 since we use fast SSDs only. With that change Postgres strongly prefers the index s_pk_action_at unless I both disable the other index and turn off bitmap heap scans.

I'm attaching the following plans:

- base_plan.txt: default costs; both indexes available

- base_plan_rpc2.txt: random_page_cost = 2; both indexes available

- inddisabled_plan_rpc2.txt: random_page_cost = 2; only s_action_at_pk available

- inddisabled_bhsoff_plan_rpc2.txt: random_page_cost = 2; enable_bitmapscan = false;  only s_action_at_pk available

A couple of questions:

- How is s_pk_action_at ever efficient to scan? Given that the highest cardinality (primary key) column is first, wouldn't an index scan effectively have to scan the entire index?

- Why does index scan on s_action_at_pk reads over 2x as many blocks as the bitmap heap scan with the same index?

- Would you expect Postgres to generally always prefer using the s_action_at_pk index over the s_pk_action_at index for this query? I realize changing the random page cost is part of what's driving this, but I still can't imagine reading the full s_pk_action_at index (assuming that's what it is doing) could ever be more valuable.

As a side note, the planner is very bad at understanding a query that happens (I realize you wouldn't write this by hand, but ORMs) when you have a where clause like:

    s.status IN ('A', 'PD', 'P') AND s.status IN ('A', 'P')

the row estimates are significantly different from a where clause with only:

    s.status IN ('A', 'P')

even though semantically those are identical.