Message-ID: <47C24AA5.50405@internode.com.au>
Date: Mon, 25 Feb 2008 15:27:09 +1030
From: Shilpa Sudhakar <ssudhakar@internode.com.au>
User-Agent: Thunderbird 1.5.0.14 (Windows/20071210)
MIME-Version: 1.0
To: Dawid Kuroczko <qnex42@gmail.com>
CC: pgsql-admin@postgresql.org
Subject: Re: WAL archiving
References: <47BE07ED.1070802@internode.com.au>
	<758d5e7f0802241701h221ddebbka852adc11f343ba6@mail.gmail.com>
In-Reply-To: <758d5e7f0802241701h221ddebbka852adc11f343ba6@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit


Thanks loads Dawid,

I'll test the process on the TEST box and note down the time it takes.


Dawid Kuroczko wrote:
> On Fri, Feb 22, 2008 at 12:23 AM, Shilpa Sudhakar
> <ssudhakar@internode.com.au> wrote:
>   
>>  Since the wal logs keep increasing, do we take the base backup every now
>>  and then so that we can delete the old log files?
>>  How often do we take a base filesystem backup keeping in mind that our
>>  systems are 24 x 7.
>>     
>
> Depends on your workload and needs.
>
> Say, if you have a policy on having ability to "go back in time" for
> at least 30 days, then
> you must keep last base backup that was done before 30 days and all
> the log files since
> then.
>
> But your primary concern should be probably recovery time after
> failure (your whole machine
> just fried ;)).  What you need to do is restore last backup and reply
> all archival WAL logs
> since then.  And assume worst possible situation (the machine fried
> just before full backup),
> for a weekly backups it means 7 days worth of WAL-log reply.
>
> Now, if your database is DML-heavy recovery won't be very fast, so you
> might want to take
> base backups more frequently.  Or better -- set up a standby server
> which will constantly
> reply the backups.
>
>  Regards,
>     Dawid Kuroczko
>
> PS: And before setting a policy, do try test recovery and note down
> how much time it took.
>