postfix-users October 2010 archive
Main Archive Page > Month Archives  > postfix-users archives
postfix-users: Re: postfix clustering

Re: postfix clustering

From: Stan Hoeppner <stan_at_nospam>
Date: Sun Oct 31 2010 - 05:39:49 GMT
To: postfix-users@postfix.org

Peter put forth on 10/29/2010 1:55 PM:

> I agree with your point.
> the above solution should work well if the active/active server
> are located in the same location.

Correct.

> for the machines in different data center, there is no guarantee of speed.

Correct.

> also, making the server run in a different data center is fail-over protection solution.

This statement above is why I'm replying a second time to this message.
 I have two critical questions regarding the apparent need for multi
site failover.

1. What are your specific failure concerns with your primary site?
Network failure? Host failure? Storage hardware failure?

2. Individual hosts can be made to me extremely reliable today both at
the hardware and software level. And with VMware ESX, for example, two
physical hosts, two virtual guests, connected to fully redundant storage
(SAN RAID with dual controllers, PSUs, etc) gives you total
active/active and/or active/passive failover depending on how you set it
up. Even with a single guest server, if one physical host dies, VMware
HA will automatically boot the downed guest on the remaining ESX host in
a matter of seconds. If you want active/active failover you simply run
two guests as an IMAP cluster with a shared SAN LUN formatted with GFS2.
 Though really, if you have ESX and the hardware setup I mention,
there's little need to build a cluster atop it. The only thing a
cluster in this scenario might help is if you had massive corruption of
the root filesystem of one of the two IMAP server guests causing it to
go down.

3. With the hardware/software failure modes covered in #2, your only
remaining concern is network availability. In this case, simply have
another (backup) internet connection installed at your facility,
preferably from a different provider.

You could choose to use a FLOSS solution instead of VMware ESX and
possibly get the same results for less money. I'm not familiar with
whether or not the FLOSS hypervisors offer anything like Vmotion and HA.
 They probably do by now.

Maybe a good question for you to ask of the members of this list at this
point is:

How many OPs here run with a multi site IMAP cluster setup with a
physically distributed mail store, either via replication or a cluster
filesystem over a wide area network?

The answers you get may be quite enlightening.

-- Stan