You can do HA from the OS level:
Have a SCSI disk array shared between two computers
Use a journalling filesystem on this array
Have a "serial kill switch"
Have a program that checks if each server is alive.
If the primary goes down, the secondary will:
a) Shut off the power to the primary
b) Mount the disks
c) Start necessary services
d) Do IP takeover of the main server's IP address
The "serial kill switch" is necessary when disk arrays are being shared,
so that two computers don't inadvertantly access the disk array at the
same time. That way, if the problem is just a comm problem between the
computers, the primary will be killed off completely, and not be able to
cause problems.
There are packages that automate parts of it, and if this is of interest
to you, I can find some good articles on it. Personally, I've found that
using a stable operating system, hardware with hotswappable drivers/power
supplies, power coming in from two different circuits, and a UPS, will
give you almost as much uptime, with less difficulty and cost.
Jon
On Mon, 25 Aug 2003, Thomas [iso-8859-1] K�hler wrote:
> Hi,
> Just having read http://developer.postgresql.org/todo.php - seems
> like quite a few features for high availibility are still
> missing. So - what's the recommended way to get some kind of HA
> up and running using PostgreSQL at the moment? How much HA will I be
> able to get?
>
> Please Cc me on Replies (I'm not subscribed to this list and
> might miss something).
>
> Bye,
> Thomas
>
> --
> Thomas K�hler Linux without limits: http://linux.s390.org/
> Millenux GmbH http://www.millenux.de/ thomas.koehler@millenux.de
> Lilienthalstra�e 2 phone: +49.711.88770.300
> D-70825 Stuttgart-Korntal fax: +49.711.88770.349
>
