Hi,
I was looking at this patch, and I was specifically curious about
how this works with prepared statements. The way the patch
works now is that it determines if the statement is to be sampled
at post_parse_analyze time which could lead to unexpected
behavior.
Let's take an example below in which the
pg_stat_statements.sample_rate is set to 0 ( to mimic
some sampling rate < 1 in which this query does not
get sampled ). At that point, all subsequent executions
of the statement will not get tracked at all. Is this
what is expected for prepared statements? My concern
is we will even lose more stats than what a user
may expect.
This of course will not be an issue for simple query.
postgres=# set pg_stat_statements.sample_rate = 0;
SET
postgres=# select pg_stat_statements_reset();
pg_stat_statements_reset
-------------------------------
2025-01-06 11:45:23.484793-06
(1 row)
postgres=# SELECT $1 \parse stmt
postgres=#
postgres=# \bind_named stmt 1 \g
?column?
----------
1
(1 row)
postgres=# \bind_named stmt 1 \g
?column?
----------
1
(1 row)
postgres=# set pg_stat_statements.sample_rate = 1;
SET
postgres=# \bind_named stmt 1 \g
?column?
----------
1
(1 row)
postgres=# \bind_named stmt 1 \g
?column?
----------
1
(1 row)
postgres=# SELECT query, calls FROM pg_stat_statements;
query | calls
-------+-------
(0 rows)
Regards,
Sami Imseih
Amazon Web Services (AWS)
