Sorry, I forgot to include the list in the to...
Any comments would be greatly appreciated, esp. in regard to nfs locking...


Jeff,

You make a very valid point, however this can be overcome in many ways.
   I am currently researching this for my employer.  Here's a general
overview of how I'm planning on designing our network/servers.

--------------------------- <- external net
        |           |
       director1    director2
        |           |
--------------------------- <- internal net
      |         |         |
     qmail1    qmail2    qmail3
      |         |         |
       \_______ |  ______/
               \|/
                |
--------------------------- <- gigabit 'nfs' net with backup 100Mb net
         |            |
       NFS1         NFS2


I'm looking at using LVS (Linux Virtual Server) to handle the load
balancing/clustering.  We'll be using the cluster for www/pop3/dns/etc
as well.  LVS will allow us to add machines dynamically.

We'll be using qmail/ldap/ldap-control and 2 (or more) LDAP servers for
qmail configuration.  I've also considered having the cluster servers
boot off of CD and use a single disk for /tmp and the queue.

I haven't fully researched the NFS servers yet, but here is my idea.
The developer that made ReiserFS also wrote/is writing DRBD, which is
capable of doing network mirroring.  I am planning to use 2 e450s that I
have with Samba (for WinXXXX servers), NFS (mirrored with DRBD), and
heartbeat software to control the failover between the 2 servers.

There are 2 obvious alternatives to the NFS solution that I mentioned
though.  NetAPP makes an appliance (Filer) that can handle >6TB (RAID5)
storage which provides for (according to some friends of mine that use
them in a 75+ e6000/45+ win2k env) 5 9s of reliability.  I don't have
prices, but I've heard that one Filer can run >$80k.  The other option
is a software solution (clustering/replication/failover) from Veritas,
which comes highly recommended.

DISCLAIMER:  I don't work for any of the above mentioned companies.

Would anyone be interested in helping me develop a FAQ for this?

Cheers,

Mike

Jeff Palmer wrote:

 > And if your NFS server goes down,  both servers are useless.  In which
 > case,  what was the point of having a backup server again?
 >
 > Jeff Palmer
 > [EMAIL PROTECTED]
 >
 > At 09:57 AM 6/13/01 +0300, you wrote:
 >
 >> Alternatively you can run two SMTP servers and one POP server. Do 
NAT for
 >> the two and export the partition with Maildirs(at the pop server) to the
 >> SMTP servers through NFS. The two servers seem to be one to the outside
 >> world. NFS can be insecure though.
 >>
 >> Joe.
 >> ----- Original Message -----
 >> From: "Henning Brauer" <[EMAIL PROTECTED]>
 >> To: <[EMAIL PROTECTED]>
 >> Sent: Saturday, June 09, 2001 4:15 AM
 >> Subject: Re: backup mail server help
 >>
 >>
 >> > On Fri, Jun 08, 2001 at 04:33:49PM -0700, Hank Wethington wrote:
 >> > > What I'd like to accomplish is if Server A is unavailable, then mail
 >> goes to
 >> > > server B. Once A is back up, server B sends the mail back to
 >> server A.
 >> Does
 >> >
 >> > On server B, add all domains in question to rcpthosts, but NOT to
 >> local
 >> or
 >> virtualdomains. That's it ;-))
 >> >
 >> > --
 >> > * Henning Brauer, [EMAIL PROTECTED], http://www.bsws.de *
 >> > * Roedingsmarkt 14, 20459 Hamburg, Germany               *
 >> > Unix is very simple, but it takes a genius to understand the
 >> simplicity.
 >> > (Dennis Ritchie)
 >> >

Reply via email to