[CentOS] Server Replication: CenOS, coda? lustre?

bryan davis bdavis at skymailer.skyauction.com
Fri Sep 16 14:42:39 UTC 2005


----- Original Message ----- 
From: "Manuel BERTRAND" <manuel.bertrand at lif.univ-mrs.fr>
To: "CentOS mailing list" <centos at centos.org>
Sent: Friday, September 16, 2005 7:10 AM
Subject: [CentOS] Server Replication: CenOS, coda? lustre?


> Hi list,
> our lab is divided in two sites and we would like to have a clone of the 
> mail server on the second site for redundancy only (used in case of a lost 
> of connectivity with the primary mail server, with the MX entries in the 
> DNS the clone will receive the mails instead). The problem(s) comes with 
> the mail partition who must be replicated and synchronized over the time 
> (and over the net) to the cloned mail server, so the users can still 
> access their mails while we deal with the primary server. My search leads 
> me to the coda FS (and maybe the lustre solution, but we only use 
> opensource soft so it may only be the one year old version - which is 
> open...).  Is there on the list people who are doing such things and if 
> you are, can you give me some clues? What I think is something like a 'one 
> read - multiple writes' FS who can handle the flow of mails and securely 
> replicates it to the clone server. Do we have no choice but to tolerate 
> the lost of some emails when the  primary server stops responding? btw is 
> it realistic?
> Thanks,
> Manuel
>



I've been looking into a similar configuration for our mail/DNS servers. One 
option I've considered is http://www.linuxha.net/index.pl . They've got a 
framework for using DRBD to maintain a replicated disk where only one node 
can be writing or reading from it at a time.  It doesn't have as many 
features as Coda or Lustre, but it seems far simpler to configure and 
administer.


Solving the inbound message problem isn't too bad as a standard backup MX 
configuration will accept mails and forward them to the primary once it 
re-appears, but the user's mail stores will be unavailable as long as the 
primary is down.

I welcome other's thoughts on the matter. 




More information about the CentOS mailing list