Re: Overload after some minutes, please help!

From: "Joshua D(dot) Drake" <jd(at)commandprompt(dot)com>
To: Peter Bauer <peter(dot)m(dot)bauer(at)gmail(dot)com>
Cc: pgsql-general(at)postgresql(dot)org
Subject: Re: Overload after some minutes, please help!
Date: 2006-10-21 16:25:51
Message-ID: 453A4A0F.1090103@commandprompt.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general


> All other loadtests (no locking or no indexing) ended up in very high
> load and an unusable system after max. one hour because of the very
> long running sub-SELECT of the DELETE statement.
>
> So i think that sometimes there were deadlocks between these 3
> statements which were detected and reported by Postgre (not sure if it
> could be resolved). This should be solved by locking the whole table.
> Additionally the sub-SELECT took so lang that vacuum couldnt clean up
> the dead rows caused by the UPDATEs and the next runtime of it was
> extremely high which lead to a unrecoverable situation because there
> was constant load.
>
> Is this a reasonable assumption or impossible nonsense?

Sounds more likely to me, since you didn't originally have indexes that
you were getting a long seqscan for your DELETE statement which was
running while the sub-select was waiting.

If you were running the DELETE statement multiple times during a load
test you likely ran into a table bloat issues because of all the dead rows.

Joshua D. Drake

>
> thx,
> Peter
>
> 2006/10/21, Peter Bauer <peter(dot)m(dot)bauer(at)gmail(dot)com>:
>> Hi,
>>
>> we had these problems with Version 7.4.7, you can find the old thread
>> here:
>> http://archives.postgresql.org/pgsql-general/2006-09/msg00079.php
>>
>> br,
>> Peter
>>
>> 2006/10/21, Chris Mair <chrisnospam(at)1006(dot)org>:
>> >
>> > > its just a vacuumdb --all. We already learned that full vacuums are
>> > > evil because the database was carrupted after some time.
>> >
>> > Wait a sec...
>> > vacuum full maybe evil in the 'locks stuff and takes long to
>> run'-sense,
>> > but it should definitly NOT corrupt your database.
>> >
>> > Are you sure there's no issues on the hardware / system administration
>> > side of things?
>> >
>> > Bye, Chris.
>> >
>> >
>> >
>> >
>> >
>> >
>> >
>>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 2: Don't 'kill -9' the postmaster
>

--

=== The PostgreSQL Company: Command Prompt, Inc. ===
Sales/Support: +1.503.667.4564 || 24x7/Emergency: +1.800.492.2240
Providing the most comprehensive PostgreSQL solutions since 1997
http://www.commandprompt.com/

Donate to the PostgreSQL Project: http://www.postgresql.org/about/donate

In response to

Browse pgsql-general by date

  From Date Subject
Next Message Marek Lewczuk 2006-10-21 17:51:35 8.2 release schedule
Previous Message Peter Bauer 2006-10-21 15:44:36 Re: Overload after some minutes, please help!