[H-GEN] Network Failover stuff

David de Groot david-humbug at viking.org.au
Thu Jun 3 08:24:22 EDT 2004


On 03/06/2004, at 9:52 PM, Michael Anthon wrote:

> [ Humbug *General* list - semi-serious discussions about Humbug and    
>  ]
> [ Unix-related topics. Posts from non-subscribed addresses will 
> vanish. ]
>
> I have been asked today to investigate network redundancy type stuff 
> for one of my servers.  The IT bods in our main office (Sydney) are 
> rehashing the main routers and DMZ configuration.  The new config will 
> have 2 Cisco 3750s.  I need to hook the built in GigE NICs to the 
> routers (one each obviously).
> The server is an HP DL380 running Debian.  My initial reaction was to 
> look at the NIC bonding stuff but after reading up a bit I'm not sure 
> that's the right tool for this job.  The initial idea that was floated 
> to me was to do some sort of IP failover thing using heartbeat tests 
> over the network.  Ideally they would prefer that I only require one 
> IP address (each machine is on it's own /30 subnet, one address for 
> the machine and one for the router) but this isn't a strict 
> requirement.
>
> At this stage I'm really just looking for ideas on where to look and 
> what sort of approach people would recommend and thought I would fire 
> this off while I continue my research.

What exactly are you trying to achieve ?

If the routers are your gateway out to the internet or somesuch and 
you're trying to configure a redundant internet (substitute lan if 
that's what you're after) connection then the easiest (although 
possible more expensive) option is as follows.

Configure HSRP on the routers. Put the inside interfaces of the 
routers, the HSRP group and the server in the same vlan off a switch. 
Point your server's default gateway at the HSRP address. Thus when one 
router goes down, the other will take over and your server won't be any 
the wiser.

Obviously though, if you're connecting directly to the routers, then 
this isn't going to work, and instead you'll have to do something a 
little more complicated. One possibility is weighted static routes on 
the server, although this seems to be inconsistent between OSs (for 
instance, and I know it doesn't apply here, but the Cisco Pix IOS 
doesn't do weighted routes - you can add them, they just don't work).

Dave





More information about the General mailing list