pg_stat_statements opening claim is potentially misleading

Started by PG Bug reporting formalmost 8 years ago2 messagesdocs
Jump to latest
#1PG Bug reporting form
noreply@postgresql.org

The following documentation comment has been logged on the website:

Page: https://www.postgresql.org/docs/10/static/pgstatstatements.html
Description:

At the top of the pg_stat_statements docs page[1]https://www.postgresql.org/docs/10/static/pgstatstatements.html it states:

"The pg_stat_statements module provides a means for tracking execution
statistics of all SQL statements executed by a server."

This is not strictly true. In cases where large numbers of queries are
cancelled, such as by `statement_timeout` on a loaded server, they are not
in fact recorded by pg_stat_statements. This has been brought up on the
mailing list [2]/messages/by-id/20171112222811.1464.28388@wrigleys.postgresql.org and has been stated as the intended behavior. This can be
surprising particularly if cancelled queries account for a large amount of
server execution time in pathological cases.

Perhaps a simple change such as this would help clarify this:

"The pg_stat_statements module provides a means for tracking execution
statistics of all SQL statements successfully executed by a server."

Or an explicit callout that cancelled queries do not count regardless of how
much execution time they actually use.

-Casey

[1]: https://www.postgresql.org/docs/10/static/pgstatstatements.html
[2]: /messages/by-id/20171112222811.1464.28388@wrigleys.postgresql.org
/messages/by-id/20171112222811.1464.28388@wrigleys.postgresql.org

#2legrand legrand
legrand_legrand@hotmail.com
In reply to: PG Bug reporting form (#1)
Re: pg_stat_statements opening claim is potentially misleading

The same should be noted for auto_explain as described here

www.postgresql-archive.org/Statements-that-time-out-are-not-considered-here-td6006753.html

As you, I was quite confused that cancelled queries where not taken into
account,
I first consider total_time, and mean_time in a second step (so durations of
timeout statements doesn't seem a problem here to me).

Here is an other discussion that could be interesting:
www.postgresql-archive.org/pg-stat-statements-how-to-catch-non-successfully-finished-statements-td6018801.html

Regards
PAscal

--
Sent from: http://www.postgresql-archive.org/PostgreSQL-docs-f2165602.html