[prev in list] [next in list] [prev in thread] [next in thread] 

List:       pgsql-performance
Subject:    Re: [PERFORM] Curious about dead rows.
From:       Jean-David Beyer <jeandavid8 () verizon ! net>
Date:       2007-11-14 21:51:44
Message-ID: 473B6DF0.8020206 () verizon ! net
[Download RAW message or body]

Alvaro Herrera wrote:
> Jean-David Beyer wrote:
>> Mario Weilguni wrote:
> 
>>> Did you rollback some transactions? It will generate dead rows too - at
>>> least I think so.
>>>
>> No, and the statistics confirm this.
> 
> To recap:
> 
> - your app only does inserts

True.

> - there has been no rollback lately

True.

> - there are no updates

True

> - there are no deletes

True.
> 
> The only other source of dead rows I can think is triggers ... do you
> have any? 

No triggers at all. I have sequences that were not in the IBM DB2 version of
this stuff. But they are all done earlier, before the CLUSTER of the entire
database. Furthermore, they are only for two tables, not the ones that
attracted my notice in the first place.

> (Not necessarily on this table -- perhaps triggers on other
> tables can cause updates on this one).
> 
> Oh, rolled back COPY can cause dead rows too.
> 
The only copies I ever do are those inside dbdump -- dbrestore, and they
come after all this stuff. And they do not roll back -- though I suppose
they could in principle.

-- 
  .~.  Jean-David Beyer          Registered Linux User 85642.
  /V\  PGP-Key: 9A2FC99A         Registered Machine   241939.
 /( )\ Shrewsbury, New Jersey    http://counter.li.org
 ^^-^^ 16:45:01 up 22 days, 10:03, 1 user, load average: 4.20, 4.22, 4.17

---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
       choose an index scan if your joining column's datatypes do not
       match
[prev in list] [next in list] [prev in thread] [next in thread]