|
On 2/9/06, Graeme Fowler <graeme@xxxxxxxxxxx> wrote:
>
> You want the track_interface to be the *other* interface, not the same
> one. If you have "interface eth0" then the "track_interface {..." must
> be eth1, and vice-versa.
Ok, switched track interface to eth0, but still get the same loop...
> However: As you're running an SMP kernel, if you're using a P4 with
> Hyperthreading (or indeed a dual-proc system) switched on then either
> turn HT off or boot with a uniprocessor kernel. Keepalived has been
> known to have problems with SMP systems, which can be worked around by
> either doing as mentioned already to go uni-proc, *or* separate the
> VRRP and healthcheck daemons by running two separate instances with the
> right options.
I dont believe we are using HT P4s. From /proc/cpuinfo: Intel(R) Pentium(R)
4 CPU 1.80GHz. Both directors have the same setup. I think the smp was
chosen because right now we are running these as test boxes. If all goes
well though, we will get dedicated server boxes for production use (most
likely HT P4s). Since this is not a HT proc, should this still cause
conflict?
-Matt
|
