Обсуждение: Explain plan on a Select query

I’m a bit confused on an explain plan I just generated for a select statement. The plan’s first line begins with the
word“Append” followed by the usual (cost....) data. What in the world is an Append doing here? I’ve recently performed
anotheranalyze on the table, btw.  

Also, the query does not seem to be using an index, as the explain shows a pair of sequential scans, speaking of which,
thesecond sequential scan shows it’s on a table, called <table>_test. Which is a little strange, considering that does
notexist. 

The query’s where clause is on a varchar(25) field. And that field has both a btree index and a trigram index on it.
I’vetried using where =, like, and ~ in order to see if I can get it to use an index. All attempts continue to say
sequentialscan. 
—
Jay
>
Sent from my iPad

Hi John,

How many rows are in the table?  If the table is small with respect to 
row count, and the pages are not very fragmented, ie, not many pages, 
then PG is just doing the right thing here: seq scan is less expensive 
than index scan in those cases.  You can turn off enable_seqscan to just 
make sure it does use the index.

Regards,
Michael Vitale

John Scalia wrote on 3/16/2020 12:26 PM:
> I’m a bit confused on an explain plan I just generated for a select statement. The plan’s first line begins with the
word“Append” followed by the usual (cost....) data. What in the world is an Append doing here? I’ve recently performed
anotheranalyze on the table, btw.
 
>
> Also, the query does not seem to be using an index, as the explain shows a pair of sequential scans, speaking of
which,the second sequential scan shows it’s on a table, called <table>_test. Which is a little strange, considering
thatdoes not exist.
 
>
> The query’s where clause is on a varchar(25) field. And that field has both a btree index and a trigram index on it.
I’vetried using where =, like, and ~ in order to see if I can get it to use an index. All attempts continue to say
sequentialscan.
 
> —
> Jay
> Sent from my iPad
>

Well, after rebuilding the statistics on the table at least three separate times, the errant messages finally disappeared. BTW, this was on an AWS RDS instance, which I know is not the favorite of any development team member. It just threw me a bit, as the first analyze did not clear the issue.



On Mon, Mar 16, 2020 at 12:26 PM John Scalia <jayknowsunix@gmail.com> wrote:

I’m a bit confused on an explain plan I just generated for a select statement. The plan’s first line begins with the word “Append” followed by the usual (cost....) data. What in the world is an Append doing here? I’ve recently performed  another analyze on the table, btw.

A top level Append would generally be used when you use a UNION ALL, or you have inheritance.

 

Also, the query does not seem to be using an index, as the explain shows a pair of sequential scans, speaking of which, the second sequential scan shows it’s on a table, called <table>_test. Which is a little strange, considering that does not exist.

Is that the name of the table, or the alias for that table?

  "->  Seq Scan on pgbench_accounts pa_alias  (cost=0.00..527869.00 rows=20000000 width=97)"

"pgbench_accounts" is the actual table, "pa_alias" is the alias for it. 

PostgreSQL will create alias for table when it needs to, but I've never seen it do so by adding "_test" to the end.  Usually it adds "_" and an integer.

It would really help to be able to see either the plan or the query, or both.

Cheers,

Jeff