Re: pg_stat_database update stats_reset only by pg_stat_reset

it reset statistics for a single table and update the column stats_reset of
pg_stat_database.
but i think that stats_reset shoud be database-level statistics, a single
table should not update the column stats_reset.

i am monitor the xact_commit every 5 minutes, when stats_reset is reset but
ohter columns is not reset, i can't decide
if i will recount the xact_commit, because pg_stat_reset make all column to
zero. pg_stat_reset_single_table_counters
only reset the column stats_reset.

张连壮 <lianzhuangzhang(at)gmail(dot)com> 于2019年5月13日周一 下午3:30写道：

> pg_stat_reset_single_table_counters/pg_stat_reset_single_function_counters
> only update pg_stat_database column stats_reset.
> stat_reset shuld update when all the column is reset.
>
> sample:
> drop database if exists lzzhang_db;
> create database lzzhang_db;
> \c lzzhang_db
>
> create table lzzhang_tab(id int);
> insert into lzzhang_tab values(1);
> insert into lzzhang_tab values(1);
>
> select tup_fetched, stats_reset from pg_stat_database where
> datname='lzzhang_db';
> select pg_sleep(1);
>
> select pg_stat_reset_single_table_counters('lzzhang_tab'::regclass::oid);
> select tup_fetched, stats_reset from pg_stat_database where
> datname='lzzhang_db';
>
> result:
> tup_fetched | stats_reset
> -------------+-------------------------------
> 514 | 2019-05-12 03:22:55.702753+08
> (1 row)
> tup_fetched | stats_reset
> -------------+-------------------------------
> 710 | 2019-05-12 03:22:56.729336+08
> (1 row)
> tup_fetched is not reset but stats_reset is reset.
>