[mdlug] Samba and F5 for Failover?

[Wojtak, Greg]

Thanks for the insight.  Yeah, at this point I'm more concerned with just being *able* to bring everything back up and having as much of the data there as possible.  In the scenario I drew out, serverA and serverB are actually red hat cluster instances.  We've got samba failover *within* a site down pat.  Going from one data center to another, not so much.  One huge challenge I am facing is that the way our company's network is designed, our vlan's don't extend from one datacenter to another, and in fact cannot without a huge, massive overhaul.  <facepalm>  So a distributed cluster, or any other kind of transparent failover is out of the question... especially with samba.

-----Original Message-----
From: mdlug-bounces at mdlug.org [mailto:mdlug-bounces at mdlug.org] On Behalf Of Adam Tauno Williams
Sent: Wednesday, October 07, 2009 10:29 AM
To: MDLUG's Main discussion list
Subject: Re: [mdlug] Samba and F5 for Failover?

On Wed, 2009-10-07 at 10:18 -0400, Wojtak, Greg wrote:
> Thanks Rob.  I think this gives me all I need.  I am failing across
> datacenters in different cities.  The requirement is for a DR
> situation, so *immediate* failover is not a requirement.

Ok, if the clients reboot you should be OK.

> Our samba servers are on our SAN and the disk is being replicated from
> one to the other.  

Make sure you are replicating snapshots and not just the live image.
That way you can go back to retrieve good copies of the [inevitable]
trashed files.

> So it sounds like I can just keep the configs in sync (maybe put all
> the required configuration info on a small replicated volume?) and
> when our primary datacenter gets exploded (ha!) I can just fire up the
> "DR" samba server and set the F5 to send requests to that server.

Until the CIFS clients reestablish state with the new file server they
will see scads of errors.

_______________________________________________
mdlug mailing list
mdlug at mdlug.org
http://mdlug.org/mailman/listinfo/mdlug