Re: Running Postgres Daemons with same data files - Mailing list pgsql-admin

From Andrew Rawnsley
Subject Re: Running Postgres Daemons with same data files
Date
Msg-id 6CF4BD49-2B3F-11D8-96E4-000393A47FCC@ravensfield.com
Whole thread Raw
In response to Re: Running Postgres Daemons with same data files  (William Yu <wyu@talisys.com>)
List pgsql-admin
On Dec 10, 2003, at 12:25 PM, William Yu wrote:

> Bhartendu Maheshwari wrote:
>
>> Dear Hal, Frank, Oli and all,
>> I understand what you all trying to say, I know this is not good way
>> of
>> designing, but we are planning for using the database for the keeping
>> mobile transactions and at the same time we need to provided the HA
>> solutions. The one solution i derive from the discussion that uses one
>> server and multiple clients but the issue in this if the system in
>> which
>> database server was running get down the its all the way no use of HA
>> and load balancing, since without data the other one can't do
>> anything.
>
> Here's an important question. What exactly is the thinking behind your
> load balancing and HA requirements? The reason I'm asking this
> question is because there's nuances to what high available means.
>
> As an example, you've got redundant servers but they're all in the
> same server room. A fire breaks out and kills everything. Not really
> HA IMO. Or you have redundant servers in different rooms/buildings
> hooked up to a NAS unit someplace else. A mover knocks the head off
> the ceiling fire extinguisher and floods the place (I've seen this
> happen) killing the NAS device. Again, not very HA. On the otherhand,
> if all your users are housed in same building as the servers where a
> fire kills the servers and also stops your users from doing any work,
> then it's not a problem.
>
> The situation my company is in is we have users all over the U.S.
> connecting to our app so to do HA, we needed to put duplicate servers
> thousands of miles away from each other. That way, an earthquake in SF
> or a terrorist attack in D.C. doesn't bring down our app. And since
> traffic was load balanced between both locations, we needed
> master-master replication which we had to code in at the app level.
>
>
> ---------------------------(end of
> broadcast)---------------------------
> TIP 2: you can get off all lists at once with the unregister command
>    (send "unregister YourEmailAddressHere" to majordomo@postgresql.org)
>

There are 2 basic things to remember about proper HA, by any definition:

1. It is hard.
2. It is expensive.

This is true even for systems which were built with HA/redundancy in
mind - Postgres wasn't. Depending on what NAS device you're talking
about, it may not
have been either (call EMC and see what the price tag is on a
redundant, HA SAN setup. If you can afford that, you can afford the
right tools).

This topic has floated around on the erserver mailing list - either you
have the budget to do HA right, or you don't. If you don't, you have to
be clear
in your specs about what is possible with the tools that you can afford
without trying to jigger inappropriate technology to do something it
really can't
do.  Trying to do something that the developers of a product say is not
possible, and selling it as HA, is not going to make you many friends
when
problems arise.


--------------------

Andrew Rawnsley
President
The Ravensfield Digital Resource Group, Ltd.
(740) 587-0114
www.ravensfield.com


pgsql-admin by date:

Previous
From: William Yu
Date:
Subject: Re: Running Postgres Daemons with same data files
Next
From: John Gibson
Date:
Subject: Re: Running Postgres Daemons with same data files