Re: Repercussions of killing a vacuum - Mailing list pgsql-novice

From s anwar
Subject Re: Repercussions of killing a vacuum
Date
Msg-id 3e3c86f90512190821u5e1f0d17s1857c4208a9bc78@mail.gmail.com
Whole thread Raw
In response to Repercussions of killing a vacuum  (s anwar <sanwar@gmail.com>)
List pgsql-novice
Good Day,

I had about four tables in my database. One of the tables made the bulk of data. It was 650GB in size. The rest of the tables were on the order of few hundred MBs if that. My 650GB table had a composite primary key and an additional unique index defined on it. The data was clustered on the index defined by the composite primary key. I ran VACUUM VERBOSE FULL (no ANALYZE) on the database. It did not finish running in over five days. The task that seemed to be the slowest seemed to be the index build.

If I manually issue a CREATE INDEX on this table, it usually takes over a day to finish.

Hope that answers your questions.

Regards,
Saadat.

On 12/15/05, DANTE ALEXANDRA <ALEXANDRA.DANTE@bull.net> wrote:
Hello,

I realize a benchmark with the release 8.1, with 300GB of data.
One of the tables contains near 300GB of data and I wonder if I have to
launch a VACUUM FULL ANALYZE on it, which could run during days.

Could you tell me that command you launch : VACUUM or VACUUM FULL
ANALYZE  and the size of the table used Your answer can help me to
estimate the execution time on 300GB.

Thank you for your answer.
Regards,
Alexandra DANTE

s anwar a écrit :

> I have a vacuum running for many days in postgres 8.1. Could the
> vacuum process be killed without causing any serious damage to the
> data. From the looks of the data directory, it seems to be rebuilding
> index files, but at a very slow speed.
>


pgsql-novice by date:

Previous
From: "Luis Silva"
Date:
Subject: pg_config not found
Next
From: Sean Davis
Date:
Subject: Re: COPY command, linking foreign keys