Yes it's ok to have to couriers writing to one files system - yes - rsync
is a good option to mirror things. Courier users
are not nessecarily HA users, and it's not an application issue - its an
implementation issue for structured HA systems where courier forms a
component.
m/
One of my options would be to provide a NFS
mountpoint that I ca provide via another cluster file
system.
Is it ok to have two courier boxes writing in to
the same NFS mounted disk ?
If I undertsand the Linux world correct
the clustered file system is the key to build bullet proof solutions. Or is
there a way to have two courier boxes replicating each other, via rsync for
example ? I do know this is probably off topic, but still, having to deal with
courier as an application, how does people solve the problem ?
/Lars
----- Original Message -----
Sent: Saturday, July 12, 2003 9:04
PM
Subject: RE: [courier-users] Disaster
tolerant IMAP and SMTP ?
You've got a few things to look into Lars - this isn't really a
courier question...
You need to look at distributed replicated storage systems (afs,
coda, or simply replicated NFS...) which are all off
topic....
Then you need to decide your method for connection failover - DNS
with recrd updates when a node fails is probably your best bet if you are
operating on different networks... you can expect your data transfer volume
to be doubled though - as ever file that is created or moved on one system
has to be transfered to the other system...
m/
The question is about how to create a
distributed imap-server with a multiple site storage.
/Lars
----- Original Message -----
Sent: Saturday, July 12, 2003 5:51
PM
Subject: Re: [courier-users]
Disaster tolerant IMAP and SMTP ?
On Saturday, Jul 12, 2003, at 07:27 US/Central, Lars
Holmstr�m wrote:
I consider building a
disaster tolerant mailsystem for in and outgoing mail./smaller>/fontfamily>
I already
have two ISPs and MX records that direct incoming mailtraffic
properly./smaller>/fontfamily>
The
problem seems to be the IMAP./smaller>/fontfamily>
Does
any one have a suggestion for how to do this ?/smaller>/fontfamily>
/Lars/smaller>/fontfamily>
Ps.
Disaster tolerant in my view is when I can afford to loose one
computerroom completely and the service from a users perspective is
still available. Ds./smaller>/fontfamily>
Is
your question about how to initially create a redundant IMAP data store
across your server rooms? Or how (after it's all created) to let users
reach it automatically, regardless of which computer room is still alive
after a failure?
I.e. Server implementation vs. DNS
implementation? (...or
both?)
-jab