MIME-Version: 1.0
From: Sheryl Prabhu David <nestor.ssn@gmail.com>
Date: Tue, 12 Mar 2024 13:00:57 +0530
Message-ID: 
 <CAKxBfWFZZO7MnvFVrJeX7bCJoXbVeeDY0Q2dJkKS8hs4ucYvXA@mail.gmail.com>
Subject: Nested loop behaviour with pg_stat_activity
To: pgsql-sql@lists.postgresql.org
Content-Type: multipart/alternative; boundary="0000000000003d4e99061371a4c8"
Archived-At: 
 <https://www.postgresql.org/message-id/CAKxBfWFZZO7MnvFVrJeX7bCJoXbVeeDY0Q2dJkKS8hs4ucYvXA%40mail.gmail.com>
Precedence: bulk

--0000000000003d4e99061371a4c8
Content-Type: text/plain; charset="UTF-8"

Hi,

I am working on a solution to sample pg_stat_activity along the lines of
what is possible in Oracle, described as follows:

In an Oracle database using a SQL like the one below, we have the ability
to use, a series along the lines of Postgresqls generate_series() as seen
in the 'connect by' line as outer table of a forced nested loop, and
v$session(Oracles equivalent of pg_stat_activity) as the inner table, to
get many samples of v$session each second.

select
        /*+ ORDERED NO_MERGE USE_NL(g t s) */
        g.*,
        t.*,
        s.*
from
        (
        select /*+ NO_MERGE */ 1 from dual connect by level <= :v_samples
        ) g,
        (
        select /*+ NO_MERGE */ hsecs from v$timer
        ) t,
        (
        select
                /*+ NO_MERGE */
                s.sql_id,
                plan_hash_value,
                sql_exec_start,
                sql_exec_id,
                event,
                state,
                wait_class,
                row_wait_obj#,
                program,
                s.module,
                s.action,
                status
        from
                v$session s
                left outer join v$sqlarea sa on
s.sql_hash_value=sa.hash_value and s.sql_address=sa.address
        where
                type != 'BACKGROUND'
        ) s
order by
        sql_id,
        sql_exec_id,
        plan_hash_value,
        event,
        row_wait_obj#

Trying something similar in Postgres does not produce an equivalent result.
Instead of 'Number of sample' times *DIFFERENT* copies of pg_stat_activity,
we are seeing 'Number of sample' times *SAME* copy of pg_stat_activity,
unlike Oracle. MVCC and Isolation guarantees for regular tables is expected
to produce this kind of a result, but I was hoping pg_stat_activity being a
portal into internal data structures will act similar to Oracles v$session
bypassing MVCC+Isolation. I am hoping to find out if there is anyway to
force Oracle type behaviour for pg_stat_activity, please help.

SQL I have tried in Postgres:

vagrant=# set enable_material = off;
SET
vagrant=# explain
with sampled as materialized (
    select
           *
    from
    (
        select
            pid,
            datname,
            usename,
            query,
            query_start,
            wait_event_type,
            wait_event,
            state
        from
            pg_stat_activity i,
            ( select * from generate_series(1,100) ) o
        offset 0
    )
    where
        state = 'active'
)
select
    query,
    query_start,
    wait_event_type,
    wait_event,
    count(*)
from
    sampled
group by
    query,
    query_start,
    wait_event_type,
    wait_event;
                                                  QUERY PLAN
---------------------------------------------------------------------------------------------------------------
 HashAggregate  (cost=330.56..331.06 rows=50 width=112)
   Group Key: sampled.query, sampled.query_start, sampled.wait_event_type,
sampled.wait_event
   CTE sampled
     ->  Subquery Scan on unnamed_subquery  (cost=2.39..328.93 rows=50
width=268)
           Filter: (unnamed_subquery.state = 'active'::text)
           ->  Nested Loop  (cost=2.39..203.93 rows=10000 width=268)
                 ->  Hash Left Join  (cost=2.38..3.93 rows=100 width=268)
                       Hash Cond: (s.usesysid = u.oid)
                       ->  Hash Left Join  (cost=1.05..2.32 rows=100
width=208)
                             Hash Cond: (s.datid = d.oid)
                             ->  Function Scan on pg_stat_get_activity s
 (cost=0.00..1.00 rows=100 width=148)
                             ->  Hash  (cost=1.02..1.02 rows=2 width=68)
                                   ->  Seq Scan on pg_database d
 (cost=0.00..1.02 rows=2 width=68)
                       ->  Hash  (cost=1.15..1.15 rows=15 width=68)
                             ->  Seq Scan on pg_authid u  (cost=0.00..1.15
rows=15 width=68)
                 ->  Function Scan on generate_series  (cost=0.00..1.00
rows=100 width=0)
   ->  CTE Scan on sampled  (cost=0.00..1.00 rows=50 width=104)
(17 rows)

Could not get the nested loop order correct with this.

vagrant=# explain
with sampled as (
    select
           *
    from
    (
    with o as (
        select
            pid,
            datname,
            usename,
            query,
            query_start,
            wait_event_type,
            wait_event,
            state
        from
            pg_stat_activity
        offset 0
    )
    select
        o2.*
    from
        o,
        o as o1,
        o as o2
        offset 0
    )
    where
        state = 'active'
)
select
    query,
    query_start,
    wait_event_type,
    wait_event,
    count(*)
from
    sampled
group by
    query,
    query_start,
    wait_event_type,
    wait_event;
                                                              QUERY PLAN
---------------------------------------------------------------------------------------------------------------------------------------
 GroupAggregate  (cost=33015.13..33139.02 rows=4889 width=112)
   Group Key: unnamed_subquery.query, unnamed_subquery.query_start,
unnamed_subquery.wait_event_type, unnamed_subquery.wait_event
   ->  Sort  (cost=33015.13..33027.63 rows=5000 width=104)
         Sort Key: unnamed_subquery.query, unnamed_subquery.query_start,
unnamed_subquery.wait_event_type, unnamed_subquery.wait_event
         ->  Subquery Scan on unnamed_subquery  (cost=3.93..32707.93
rows=5000 width=104)
               Filter: (unnamed_subquery.state = 'active'::text)
               ->  Nested Loop  (cost=3.93..20207.93 rows=1000000 width=268)
                     CTE o
                       ->  Hash Left Join  (cost=2.38..3.93 rows=100
width=268)
                             Hash Cond: (s.usesysid = u.oid)
                             ->  Hash Left Join  (cost=1.05..2.32 rows=100
width=208)
                                   Hash Cond: (s.datid = d.oid)
                                   ->  Function Scan on
pg_stat_get_activity s  (cost=0.00..1.00 rows=100 width=148)
                                   ->  Hash  (cost=1.02..1.02 rows=2
width=68)
                                         ->  Seq Scan on pg_database d
 (cost=0.00..1.02 rows=2 width=68)
                             ->  Hash  (cost=1.15..1.15 rows=15 width=68)
                                   ->  Seq Scan on pg_authid u
 (cost=0.00..1.15 rows=15 width=68)
                     ->  Nested Loop  (cost=0.00..203.00 rows=10000 width=0)
                           ->  CTE Scan on o  (cost=0.00..2.00 rows=100
width=0)
                           ->  CTE Scan on o o1  (cost=0.00..2.00 rows=100
width=0)
                     ->  CTE Scan on o o2  (cost=0.00..2.00 rows=100
width=136)
(21 rows)

With this I am able to get the nested loop order correct where the alias o2
from which we are taking the results is the inner table in join order, but
still we only get multiple copies of one state from pg_stat_activity.

Please suggest alternatives if any.

Regards
Prabhu David

--0000000000003d4e99061371a4c8
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: base64
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--0000000000003d4e99061371a4c8--