Re: ANALYZE after CREATE TABLE AS SELECT... - Mailing list pgsql-general

From David Steele
Subject Re: ANALYZE after CREATE TABLE AS SELECT...
Date
Msg-id 54EFC5E7.9050002@pgmasters.net
Whole thread Raw
In response to Re: ANALYZE after CREATE TABLE AS SELECT...  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: ANALYZE after CREATE TABLE AS SELECT...  ("Semyon Reyfman" <sreyfman@therealauthority.com>)
List pgsql-general
On 2/26/15 2:05 PM, Tom Lane wrote:
> "Semyon Reyfman" <sreyfman@therealauthority.com> writes:
>> When I create a table with "CREATE TABLE name AS SELECT." statement and
>> immediately afterward use this new table in a query does it make sense to
>> run ANALYZE on the new table in between?
>
> Yes.

Yes.  And to be more specific - vacuum cannot see tables that have been
created in a transaction, so you should always analyze tables that you
intend to read in the same transaction where you created them.

Even if you do commit before reading there is a race condition with
vacuum, so it's best to analyze.

Bonus tip: the same is true for any temp tables you might create.  More
so, since vacuum will never see them at all.

--
- David Steele
david@pgmasters.net


Attachment

pgsql-general by date:

Previous
From: Tim Uckun
Date:
Subject: Re: Triggers and scalability in high transaction tables.
Next
From: David Steele
Date:
Subject: Re: ANALYZE after CREATE TABLE AS SELECT...