On Fri, 2012-05-11 at 05:40 -0700, John Doe wrote:

Hi,

right now I am using only one external server as a gateway for the internal servers. I would like to enable a fail-over on a second server. To implement the floating VIP, should I use heartbeat+pacemaker? Or is there something more "lightweight"? Basically, I just need server2 up the VIP when server1 is down, and server2 down the VIP when server1 is back up (or server 1 does not up the VIP seeing server2 has it already).

Thx, JD

Heartbeat is a pretty lightweight way of doing this. With any failover/VIP setup there is always the possibility of a split-brain but for the most part I haven't seen this happen in my environment.

Also, use IPaddr2 as the resource type as it will do an gratuitous arp, significantly decreasing the failover time.

eg. [haresources file]

node.fqdn IPaddr2::10.10.10.10/24/eth1

Substitute your IP, mask and network interface.

There are some options you can set as well that tell it not to fail back automatically, just so you can check the box that failed before pushing services back to it.

Tait

On 05/11/2012 02:40 PM, John Doe wrote:

Hi,

right now I am using only one external server as a gateway for the internal servers. I would like to enable a fail-over on a second server. To implement the floating VIP, should I use heartbeat+pacemaker? Or is there something more "lightweight"? Basically, I just need server2 up the VIP when server1 is down, and server2 down the VIP when server1 is back up (or server 1 does not up the VIP seeing server2 has it already).

If you only need IP handling then you can use keeplived which is more lightweight than pacemaker but can only deal with networking and nothing else.

Hearbeat isn't developed anymore so if you want to go the pacemaker route you probably want to go for corosync+pacemaker instead.

Regards, Dennis