RFC: failover configuration

john hernandez jph at sjrwmd.state.fl.us
Thu Apr 22 14:38:29 GMT 1999

I've recently been asked to investigate the possibility of using our 2 Samba 
fileservers (strictly) in a manually-enabled symmetric failover configuration 
for read-only file shares.

Using rsync :) I've managed to replicate the predominantly static filesystems 
between our two Unix servers.  Now, I'm testing a script which starts a second 
set of nmbd and smbd procs (in daemon mode) by specifying an alternate 
configuration file.  My initial tests seem to indicate that this may actually 

At our site, we use Version 1.9.18p10, server (proxy) authentication to an NT 
PDC, WINS, and limit nmbd to non-master-browser/non-WINS-server roles.  We're 
not concerned with print services, just file services.

The changes to my original smb.conf as well as the alternate smb.failover.conf 
files on both servers are summarized below:

bind interfaces only = true
interfaces = primary.interface.ip.addr/mask

smb.failover.conf  -- same as smb.conf except for:
netbios name = other_hostname
interfaces = virtual.interface.ip.addr/mask
replaced primary shares with directories of replicated data

When a primary server fails (disk crash, system board, whatever), I can login to 
the other system and run a script which will ideally come to the rescue:

1) ifconfig a virtual interface (such as eth0:1) with the same IP addr. as the 
failed server
2) nmbd -D -s smb.failover.conf
3) smbd -D -s smb.failover.conf

and (hopefully?)... viola!

How will smb clients react to this?  Will the sudden ethernet (MAC) address 
change cause any hardship to NT clients?  How about lock (var) directories, can 
they be shared between the two sets of server processes?  Any other caveats that 
I'm overlooking?  Am I in violation any major Samba dogma?  Any comments would 
be greatly appreciated.

Please, also reply by direct email, as i am not a regular subscriber to this 

TIA, John

More information about the samba mailing list