|
Christian Joelly <christian.joelly@xxxxxxxxx> wrote:
>
> i have troubles with my heartbeat setup for director failover:
First, it' a better idea to ask Heartbeat related questions on the
Heartbeat mailing list rather than on the LVS list, though some people here
have at least some experience with Heartbeat.
> i use a serial null modem cable on /dev/ttyS0 and udp port 694 on eth0
> on both nodes:
[SNIP]
> ha-log from node 1: (main director)
> heartbeat: 2001/11/27_18:05:35 info: Configuration validated. Starting hear=
> tbeat.
> heartbeat: 2001/11/27_18:05:36 error: Cannot open /proc/ha/.control: No suc=
> h file or directory
Because of this message I assume that you are running a rather outdated
version of Heartbeat. If so I strongly recommend to update your Heartbeat.
[SNIP]
> but: when i pull the network cable from eth0:node1, the backup node 2
> never takes the resource as defined in the haresources file:
>
> edvlvsd1 IPaddr::10.17.160.30/16
>
> (same content on both machines!)
>
> when i start a cat </dev/ttyS0 on the backup node 2, then i can see that
> the node 1 still sends up msgs to the node 2 via the nullmodem cable...
> i dont know why the node 2 doesnt take over the resource from node 1...
> maybe im looking in the wrong direction?
It doesn't take over because you have a redundant heartbeat path: the
serial connection. Both nodes can still communicate via this path so there
won't be a failover (btw. this is a FAQ).
What you need is a additional application like mon
(http://www.kernel.org/pub/software/admin/mon/html/), which monitors your
resources (where network connectivity is one from) and take appropiate
steps if a resource fails.
Cheers,
Juri
--
Juri Haberland <juri@xxxxxxxxxxxxxx>
|
