That sounds great. Thanks for clearing that up.
Wes Sheldahl
Tom Lane <tgl%sss.pgh.pa.us@interlock.lexmark.com> on 11/15/2001 05:52:00 PM
To: "Wesley_Sheldahl/Lex/Lexmark.LEXMARK"@sweeper.lex.lexmark.com
cc: pgsql-general%postgresql.org@interlock.lexmark.com (bcc: Wesley
Sheldahl/Lex/Lexmark)
Subject: Re: [GENERAL] Idle in transaction ????
wsheldah@lexmark.com writes:
> If vacuum in 7.2 skips tables it can't lock,
It does not.
The real change in 7.2 is that vacuum requires only an ordinary writer's
lock on the table, not exclusive lock. Since a write lock doesn't
conflict with read or write locks (basically it only conflicts with
schema-changing operations) we expect that vacuum will run concurrently
with most ordinary database operations.
regards, tom lane
