From: | Lars Heidieker <lars(at)heidieker(dot)de> |
---|---|
To: | Gábor Farkas <gabor(at)nekomancer(dot)net> |
Cc: | Simon Riggs <simon(at)2ndquadrant(dot)com>, "Joshua D(dot) Drake" <jd(at)commandprompt(dot)com>, pgsql-general(at)postgresql(dot)org |
Subject: | Re: vacuum, dead rows, usual solutions didn't help |
Date: | 2008-01-10 11:05:59 |
Message-ID: | 14573F45-4B72-43DD-884F-A704A2A73C38@heidieker.de |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On 10 Jan 2008, at 11:18, Gábor Farkas wrote:
> Simon Riggs wrote:
>>> also, even if it is wrong, can an 'idle-in-transaction'
>>> connection that was opened today block the vacuuming of rows that
>>> were deleted yesterday?
>> Yes, if the rows were deleted after the connection started.
>
> to avoid any potential misunderstandings, i will summarize the
> situation:
>
> 1. the vacuum-cronjob refuses to remove dead rows since 1.jan.2008.
>
> 2. i know that no postgres-process is older than 7.jan.2008. (from
> "ps aux | grep postgres", and except the postgres-system-processes)
>
> how can this happen?
>
>
To my understanding the question how old the processes are is only
partially of interest,
if a process touches those rows in a transaction just before the
vacuum runs, it can't remove those rows.
So all you need to get in this situation is a transaction that
touches the rows in a transaction and
keeps the transaction alive before vacuum runs.
- --
Viele Grüße,
Lars Heidieker
lars(at)heidieker(dot)de
http://paradoxon.info
- ------------------------------------
Mystische Erklärungen:
Die mystischen Erklärungen gelten für tief;
die Wahrheit ist, dass sie noch nicht einmal oberflächlich sind.
-- Friedrich Nietzsche
[ Die Fröhliche Wissenschaft Buch 3, 126 ]
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)
iD8DBQFHhfwXcxuYqjT7GRYRAoKcAKCZgW/RI9rWN0/Gkd+c7F3T4WmV0gCg4Y6p
VBxOBw50HJYsHBPFUjuaPa4=
=8d+w
-----END PGP SIGNATURE-----
From | Date | Subject | |
---|---|---|---|
Next Message | Gokulakannan Somasundaram | 2008-01-10 11:13:23 | Re: Performance problem. Could it be related to 8.3-beta4? |
Previous Message | Harald Fuchs | 2008-01-10 11:04:06 | Re: count(*) and bad design was: Experiences with extensibility |