Re: How to implement backup protocol

From: "Andrus" <eetasoft(at)online(dot)ee>
To: pgsql-general(at)postgresql(dot)org
Subject: Re: How to implement backup protocol
Date: 2006-11-28 16:01:43
Message-ID: ekhmhs$1g6o$1@news.hub.org
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

> The weekly backup of the larger of the two databases produces a file that
> is about 20GB and takes about an hour and 15 minutes. I then compress it
> down to about 4 GB, which takes another hour. However, because that's a
> separate task, it doesn't impact the database server as much. (I suspect
> all that I/O slows things down a bit, but I haven't noticed any
> significant effect in my transaction time reports. That task is run during
> the slowest 4 hour period of the week, though).

My environment is a bit different. For safety, I need to create backups to
separate computer over over internet.

1. Backup computer has client category internet connection (ADSL, 600 KBit/s
download speed).

2. Query "SELECT sum( relpages * 8/1000) FROM pg_class" returns 1302 for
a database restored from backup.
So my data size seems to be approx 1 GB only.

3. Backup file size is 70 M

4. Backup client has all ports closed.

5. Server has *only* 5432 port open.

So I think that 4.5 hours which requires to create backup is because pg_dump
download the whole database (1 GB) in uncompressed format over slow
internet connection.
Compression level does not affect to this almost at all.

I think I can create backup copy fast in server computer but how to send it
to backup computer?

pg_read_file() can read only text files and is restricted only to
superusers.

How to add a function pg_read_backup() to Postgres which creates and
returns backup file with download speed ?
This problably requires implementing some file download protocol.

> BTW, if you've never actually tested your recovery capabilities, can you
> be sure they work?
> I did a full-blown test in February or March and found a few loose ends.
> And when we had to do the real thing in May (due to a power supply
> failure), there were STILL a few loose ends, but we were back online
> within 12 hours of when I started the recovery process, and half of that
> time was spent completing the setup of the 'backup' server, which I had
> been rebuilding. I'm working to lower that downtime and will be doing
> another full-blown test in January or February.

I expect that full database backup created using pd_dump does not have never
have any problems on restore.

Andrus.

In response to

Responses

Browse pgsql-general by date

  From Date Subject
Next Message Tony Caduto 2006-11-28 16:05:51 Re: Development of cross-platform GUI for Open Source DBs
Previous Message Stephen Harris 2006-11-28 16:00:35 Datafiles binary portable?