Date: Fri, 27 Jul 2001 13:28:34 +0100 (IST) From: Sascha Lucky Luck <bofh@online.ie> To: Andrew Reid <andrew.reid@plug.cx> Cc: freebsd-isp@FreeBSD.ORG Subject: Re: FreeBSD Mail Toaster CLUSTER Message-ID: <996236914.3b615e72f3c96@mail.online.ie> In-Reply-To: <996207686.1405.14.camel@percible.alfred.cx> References: <20010726065304.39908.qmail@web20106.mail.yahoo.com> <3B5FDE77.DC5FC3DB@i-clue.de> <996207686.1405.14.camel@percible.alfred.cx>
next in thread | previous in thread | raw e-mail | index | archive | help
Quoting Andrew Reid <andrew.reid@plug.cx>:
> The notion of setting up a secondary MX host is easy enough,
> but it
> doesn't seem to be enough for what I'd like to do.
>
> For example:
>
> +-----------------------+ +---------+
> | |---------------->| MX1 |
> | Load Balancing Device | +---------+
> | | +---------+
> | |---------------->| MX2 |
> +-----------------------+ +---------+
>
> If the Load Balancing Device ("LBD") is sharing out
> connections for
> ports 25 and 110 between MX1 and MX2, I want to be sure that
> if one of
> the two machines goes down, everyone can still get their
> mail.
>
> I've looked at some methods of doing this where, when accounts
> are
> created, they are equally dispersed amongst your MX farm, but
> that means
> that all users on the dead MX will have no mail access.
>
> How are people going about delivering distributed,
> redundant,
> load-balanced mail? I was thinking that some form of parallel
> delivery
> could be implemented, but that would result in the load of MX1
> and MX2
> being pretty similar (one would think).
>
> You don't want to have all your eggs in one basket, but at the
> same
> time, you do. You want to be able to get to your mail -all-
> the time,
> not just when the appropriate server is up. If accounts "joe",
> "thelma"
> and "louise" are on MX1, they still need to be able to get to
> their
> mail, even when the server is down.
>
> Hmm! Suggestions?
The file system where the user accounts live must then be shared
between the different mail servers. We mounted that from a Network
Appliance Filer cluster, so that was pretty safe to begin with.
User account information needs to either reside on the shared FS or
must be duplicated (eg. rdist) between the machines.
If you are connected to the server that failed you'll have to
reconnect. Otherwise the user will never notice a server failure.
> - andrew
HTH,
s.
----------------------------------------------------
Get your free email account at http://www.online.ie/
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-isp" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?996236914.3b615e72f3c96>