From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Pailloncy Jean-Gérard <pailloncy(at)ifrance(dot)com> |
Cc: | pgsql-performance(at)postgresql(dot)org |
Subject: | Re: Index Backward Scan fast / Index Scan slow ! |
Date: | 2004-04-12 21:23:46 |
Message-ID: | 13960.1081805026@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
[ Ah, I just got to your message with the VACUUM VERBOSE results ... ]
=?ISO-8859-1?Q?Pailloncy_Jean-G=E9rard?= <pailloncy(at)ifrance(dot)com> writes:
> dps=# VACUUM VERBOSE url;
> INFO: index "url_next_index_time" now contains 349519 row versions in
> 3534 pages
> DETAIL: 229515 index row versions were removed.
> 3071 index pages have been deleted, 2864 are currently reusable.
> CPU 0.32s/0.67u sec elapsed 76.25 sec.
Hm, this is odd. That says you've got 349519 live index entries in only
463 actively-used index pages, or an average of 754 per page, which
AFAICS could not fit in an 8K page. Are you using a nondefault value of
BLCKSZ? If so what?
If you *are* using default BLCKSZ then this index must be corrupt, and
what you probably need to do is REINDEX it. But before you do that,
could you send me a copy of the index file?
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Ken Geis | 2004-04-13 00:17:31 | Tracking down performance issue |
Previous Message | Tom Lane | 2004-04-12 21:02:47 | Re: index v. seqscan for certain values |