Re: Parallel Append subplan order instability on aye-aye

Thomas Munro Mon, 20 May 2019 16:33:15 -0700

On Mon, May 20, 2019 at 4:46 PM Tom Lane <t...@sss.pgh.pa.us> wrote:
> Thomas Munro <thomas.mu...@gmail.com> writes:
> > Here's a one-off regression test failure of a sort that commit
> > 624e440a intended to fix.
>
> Note that in the discussion that led up to 624e440a, we never did
> think that we'd completely explained the original irreproducible
> failure.
>
> I think I've seen a couple of other cases of this same failure
> in the buildfarm recently, but too tired to go looking right now.


I think it might be dependent on incidental vacuum/analyze activity
having updated reltuples.  With the attached script, I get the two
plan variants depending on whether I comment out "analyze a_star".  I
guess we should explicitly analyze these X_star tables somewhere?

-- 
Thomas Munro
https://enterprisedb.com

drop table if exists a_star cascade;

--- extracted from from create_table.sql

CREATE TABLE a_star (
        class           char,
        a                       int4
);

CREATE TABLE b_star (
        b                       text
) INHERITS (a_star);

CREATE TABLE c_star (
        c                       name
) INHERITS (a_star);

CREATE TABLE d_star (
        d                       float8
) INHERITS (b_star, c_star);

CREATE TABLE e_star (
        e                       int2
) INHERITS (c_star);

CREATE TABLE f_star (
        f                       polygon
) INHERITS (e_star);

--- extracted from create_misc.sql

INSERT INTO a_star (class, a) VALUES ('a', 1);

INSERT INTO a_star (class, a) VALUES ('a', 2);

INSERT INTO a_star (class) VALUES ('a');

INSERT INTO b_star (class, a, b) VALUES ('b', 3, 'mumble'::text);

INSERT INTO b_star (class, a) VALUES ('b', 4);

INSERT INTO b_star (class, b) VALUES ('b', 'bumble'::text);

INSERT INTO b_star (class) VALUES ('b');

INSERT INTO c_star (class, a, c) VALUES ('c', 5, 'hi mom'::name);

INSERT INTO c_star (class, a) VALUES ('c', 6);

INSERT INTO c_star (class, c) VALUES ('c', 'hi paul'::name);

INSERT INTO c_star (class) VALUES ('c');

INSERT INTO d_star (class, a, b, c, d)
   VALUES ('d', 7, 'grumble'::text, 'hi sunita'::name, '0.0'::float8);

INSERT INTO d_star (class, a, b, c)
   VALUES ('d', 8, 'stumble'::text, 'hi koko'::name);

INSERT INTO d_star (class, a, b, d)
   VALUES ('d', 9, 'rumble'::text, '1.1'::float8);

INSERT INTO d_star (class, a, c, d)
   VALUES ('d', 10, 'hi kristin'::name, '10.01'::float8);

INSERT INTO d_star (class, b, c, d)
   VALUES ('d', 'crumble'::text, 'hi boris'::name, '100.001'::float8);

INSERT INTO d_star (class, a, b)
   VALUES ('d', 11, 'fumble'::text);

INSERT INTO d_star (class, a, c)
   VALUES ('d', 12, 'hi avi'::name);

INSERT INTO d_star (class, a, d)
   VALUES ('d', 13, '1000.0001'::float8);

INSERT INTO d_star (class, b, c)
   VALUES ('d', 'tumble'::text, 'hi andrew'::name);

INSERT INTO d_star (class, b, d)
   VALUES ('d', 'humble'::text, '10000.00001'::float8);

INSERT INTO d_star (class, c, d)
   VALUES ('d', 'hi ginger'::name, '100000.000001'::float8);

INSERT INTO d_star (class, a) VALUES ('d', 14);

INSERT INTO d_star (class, b) VALUES ('d', 'jumble'::text);

INSERT INTO d_star (class, c) VALUES ('d', 'hi jolly'::name);

INSERT INTO d_star (class, d) VALUES ('d', '1000000.0000001'::float8);

INSERT INTO d_star (class) VALUES ('d');

INSERT INTO e_star (class, a, c, e)
   VALUES ('e', 15, 'hi carol'::name, '-1'::int2);

INSERT INTO e_star (class, a, c)
   VALUES ('e', 16, 'hi bob'::name);

INSERT INTO e_star (class, a, e)
   VALUES ('e', 17, '-2'::int2);

INSERT INTO e_star (class, c, e)
   VALUES ('e', 'hi michelle'::name, '-3'::int2);

INSERT INTO e_star (class, a)
   VALUES ('e', 18);

INSERT INTO e_star (class, c)
   VALUES ('e', 'hi elisa'::name);

INSERT INTO e_star (class, e)
   VALUES ('e', '-4'::int2);

INSERT INTO f_star (class, a, c, e, f)
   VALUES ('f', 19, 'hi claire'::name, '-5'::int2, '(1,3),(2,4)'::polygon);

INSERT INTO f_star (class, a, c, e)
   VALUES ('f', 20, 'hi mike'::name, '-6'::int2);

INSERT INTO f_star (class, a, c, f)
   VALUES ('f', 21, 'hi marcel'::name, '(11,44),(22,55),(33,66)'::polygon);

INSERT INTO f_star (class, a, e, f)
   VALUES ('f', 22, '-7'::int2, 
'(111,555),(222,666),(333,777),(444,888)'::polygon);

INSERT INTO f_star (class, c, e, f)
   VALUES ('f', 'hi keith'::name, '-8'::int2,
           '(1111,3333),(2222,4444)'::polygon);

INSERT INTO f_star (class, a, c)
   VALUES ('f', 24, 'hi marc'::name);

INSERT INTO f_star (class, a, e)
   VALUES ('f', 25, '-9'::int2);

INSERT INTO f_star (class, a, f)
   VALUES ('f', 26, '(11111,33333),(22222,44444)'::polygon);

INSERT INTO f_star (class, c, e)
   VALUES ('f', 'hi allison'::name, '-10'::int2);

INSERT INTO f_star (class, c, f)
   VALUES ('f', 'hi jeff'::name,
           '(111111,333333),(222222,444444)'::polygon);

INSERT INTO f_star (class, e, f)
   VALUES ('f', '-11'::int2, '(1111111,3333333),(2222222,4444444)'::polygon);

INSERT INTO f_star (class, a) VALUES ('f', 27);

INSERT INTO f_star (class, c) VALUES ('f', 'hi carl'::name);

INSERT INTO f_star (class, e) VALUES ('f', '-12'::int2);

INSERT INTO f_star (class, f)
   VALUES ('f', '(11111111,33333333),(22222222,44444444)'::polygon);

INSERT INTO f_star (class) VALUES ('f');

--- depending on whether or not a_star is analyzed, we get a different plan

--analyze a_star;
analyze b_star;
analyze c_star;
analyze d_star;
analyze e_star;
analyze f_star;

--- extracted from select_parallel.sql (and slightly modified)

set parallel_setup_cost=0;
set parallel_tuple_cost=0;
set min_parallel_table_scan_size=0;
set max_parallel_workers_per_gather=4;

explain (costs off)
  select round(avg(a)), sum(a) from a_star;

Re: Parallel Append subplan order instability on aye-aye

Reply via email to