Re: PostgreSQL in Cluster

From: Anton Nikiforov <Anton(dot)Nikiforov(at)loteco(dot)ru>
To: pgsql-general(at)postgresql(dot)org
Subject: Re: PostgreSQL in Cluster
Date: 2004-02-28 08:43:45
Message-ID: 404054C1.6010603@loteco.ru
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

> Hi everyone,
Hello!
>
> Anyone have PostgreSQL running on a cluster?
Yes. Someone :) Me :)
> I'm trying to run on a OpenMosix 2-machine cluster but the processes are
> not migrating.
The solution you are looking for is depending on what you want to get
srom it. I mean even oracle cluster solution do not allow to migrate
updates if while update is going on the hardware fail happend. It
migrate to a second node and all selects are starting from the beginning
and update and inserts are canselled (as far as i know).
So to find a solution you should know:
1. How much time your database could be unavailable (one second per year
or one hour per day)
2. Does your application could recoonect to a database if the problem arise.
3. Does your application is cappable to check data consistancy (after
insert, update whatever if fail happend)
............
1001.
As for me i answerred all that questions and have the following
configuration:
1. Two Intel SRKA4 machines with RAID1 for OS and RAID5 for databasee (i
know that RAID5 is too slow but for me it is just a solution)
2. Bouth computers connected to different power feeders via different UPSs.
3. I have 3 dedicated Gigabit ethernet connections established between
computers for data mirroring and two more for data access from the network.
4. Red Hat Linux8 installed it is a pitty thing because i'm using
FreeBSD on all other servers but DRBD is available only for linux.
5. All disks being mirrored over a network using DRBD software.
All database drives are configured to write data to remote node first
(Protocol C).
6. Hardware being monitored with heartbeat software that is configured
to send beats each second and if 2 heartbeats are missing it consider it
is as hardware mulfunction
7. Postgress migrating from one computer to another in 5 seconds.
8. There is a special daemon that controlls data consistancy.

Best regards,
Anton
> All the load is concentrating in one machine.
> And, how can I have fail over?
>
> Thank you,
>
> Luiz

In response to

Browse pgsql-general by date

  From Date Subject
Next Message Rajat Katyal 2004-02-28 09:51:25 Trigger function aborts the transaction on exception
Previous Message Rajat Katyal 2004-02-28 08:36:54 Re: value too long error