From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | shridhar_daithankar(at)persistent(dot)co(dot)in |
Cc: | pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: Doubt w.r.t vacuum |
Date: | 2003-07-28 13:44:41 |
Message-ID: | 8797.1059399881@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
"Shridhar Daithankar" <shridhar_daithankar(at)persistent(dot)co(dot)in> writes:
> 1. IIRC vacuum recovers/reuses dead tuples generated from update but can not do
> so for delete? Why?
This is not correct.
> 2. Vacuum full locks entire table, is it possible that it locks a page at a
> time and deal with it.
No. You can't compact the table by moving tuples without locking the
entire table. (For example, if we move a tuple from the end down to an
earlier page, it's quite possible that a concurrently executing
sequential scan would miss that tuple entirely. Another problem is that
we cannot truncate the table to fewer pages without locking out writers;
else we may decide that there are N empty pages, then execute ftruncate()
just after someone has put a new tuple into one of those pages.)
Non-full vacuum is designed specifically to do what can be done without
an exclusive lock.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Tom Lane | 2003-07-28 13:48:15 | Re: Warning for undefined cursor |
Previous Message | Tom Lane | 2003-07-28 13:34:41 | Re: Some macros for error field codes |