Re: New SQL counter statistics view (pg_stat_sql)

From: Haribabu Kommi <kommi(dot)haribabu(at)gmail(dot)com>
To: Dilip Kumar <dilipbalaut(at)gmail(dot)com>
Cc: vinayak <Pokale_Vinayak_q3(at)lab(dot)ntt(dot)co(dot)jp>, "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: New SQL counter statistics view (pg_stat_sql)
Date: 2017-01-18 05:15:30
Message-ID: CAJrrPGcDg2DXhNU+z1nbBBCJ3bOUQpo5BKxsd3YFiVCEwfJicQ@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Sat, Jan 14, 2017 at 2:58 AM, Dilip Kumar <dilipbalaut(at)gmail(dot)com> wrote:

> On Wed, Jan 11, 2017 at 7:47 PM, Dilip Kumar <dilipbalaut(at)gmail(dot)com>
> wrote:
> > +void
> > +pgstat_count_sqlstmt(const char *commandTag)
> > +{
> > + PgStat_SqlstmtEntry *htabent;
> > + bool found;
> > +
> > + if (!pgstat_track_sql)
> > + return
> >
> > Callers of pgstat_count_sqlstmt are always ensuring that
> > pgstat_track_sql is true, seems it's redundant here.
>
> I have done testing of the feature, it's mostly working as per the
> expectation.
>

Thanks for the review and test.

The use case for this view is to find out the number of different SQL
statements
that are getting executed successfully on an instance by the
application/user.

I have few comments/questions.
>
> --------
> If you execute the query with EXECUTE then commandTag will be EXECUTE
> that way we will not show the actual query type, I mean all the
> statements will get the common tag "EXECUTE".
>
> You can try this.
> PREPARE fooplan (int) AS INSERT INTO t VALUES($1);
> EXECUTE fooplan(1);
>
> ------
>

Yes, that's correct. Currently the count is incremented based on the base
command,
because of this reason, the EXECUTE is increased, instead of the actual
operation.

> + /* Destroy the SQL statement counter stats HashTable */
> + hash_destroy(pgstat_sql);
> +
> + /* Create SQL statement counter Stats HashTable */
>
> I think in the patch we have used mixed of "Stats HashTable" and
> "stats HashTable", I think better
> to be consistent throughout the patch. Check other similar instances.
>
> ----------------
>

Corrected.

>
> @@ -1102,6 +1102,12 @@ exec_simple_query(const char *query_string)
>
> PortalDrop(portal, false);
>
> + /*
> + * Count SQL statement for pg_stat_sql view
> + */
> + if (pgstat_track_sql)
> + pgstat_count_sqlstmt(commandTag);
>
> We are only counting the successful SQL statement, Is that intentional?
>

Yes, I thought of counting the successful SQL operations that changed the
database over a period of time. I am not sure whether there will be many
failure operations that can occur on an instance to be counted.

> ------
> I have noticed that pgstat_count_sqlstmt is called from
> exec_simple_query and exec_execute_message. Don't we want to track the
> statement executed from functions?
> -------

The view is currently designed to count user/application initiated SQL
statements,
but not the internal SQL queries that are getting executed from functions
and etc.

>>+void
>>+pgstat_count_sqlstmt(const char *commandTag)
>>+{
>>+ PgStat_SqlstmtEntry *htabent;
>>+ bool found;
>>+
>>+ if (!pgstat_track_sql)
>>+ return
>
>Callers of pgstat_count_sqlstmt are always ensuring that
>pgstat_track_sql is true, seems it's redundant here.

Removed the check in pgstat_count_sqlstmt statement and add it
exec_execute_message
function where the check is missed.

Updated patch attached.

Regards,
Hari Babu
Fujitsu Australia

Attachment Content-Type Size
pg_stat_sql_row_mode_4.patch application/octet-stream 25.9 KB

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Michael Paquier 2017-01-18 05:22:00 Re: Protect syscache from bloating with negative cache entries
Previous Message Dilip Kumar 2017-01-18 05:14:58 Re: Parallel bitmap heap scan