[DRBD-user] Connectivity question
Art Age Software
artagesw at gmail.com
Wed Dec 19 01:12:15 CET 2007
Yes, the easiest way to add an additional link for heartbeat, if you
don't have serial, is to use mcast or bcast on the other interface as
you have done. You definitely need heartbeat running over at least one
additional link other than your DRBD link.
On Dec 18, 2007 3:18 PM, Rois Cannon <rois at cobiz.com> wrote:
> If this is just your development version a serial link probably isn't
> necessary. I wanted to see how it would work so I created a serial link
> in my vmware server setup for both nodes. It's just one more
> communication path for redundancy. I am planning to use it when I
> implement in production.
> On Tue, 2007-12-18 at 16:29 -0500, Greg Haase wrote:
> > Right.
> > Actually, I found this link on Florian's blog about dopd:
> > http://fghaas.wordpress.com/2007/10/01/an-underrated-cluster-admins-companion-dopd/
> > ... Which subsequently has a link to this page:
> > http://fghaas.wordpress.com/2007/08/07/configuring-heartbeat-links/
> > ... Which describes exactly as you say. eth1, eth0 and serial. I have
> > just got mcast setup on eth0 to accompany bcast on eth1 and done the dopd
> > configs. I'm going to do some testing with this setup and see what
> > happens. I haven't explored creating a serial connection on a virtual
> > machine, but I never seem to have luck with serial connections in general
> > (thinking minicom) so I'm a little hesitant to go there.
> > > I'm doing something similar and although I'm pretty new at this I'm
> > > gathering the best setup would be to have heartbeat running over both
> > > eth0 and eth1 as well as providing a serial connection. That way if the
> > > drbd link goes down, the peer-outdater program can use one of the other
> > > heartbeat paths/links to outdate the secondary and keep heartbeat from
> > > using it.
> > >
> > > On Tue, 2007-12-18 at 10:03 -0500, Greg Haase wrote:
> > >> I'm playing around with a drbd + mysql setup, and I had some questions
> > >> about how the servers should be connected.
> > >>
> > >> I currently have two virtual machines running, each with two ethernet
> > >> ports. On each machine eth0 is Bridged - so it's connected to our
> > >> development subnet. The eth1 ports are both on a private network within
> > >> VM (VMnet5). I think that's the best way I can simulate a
> > >> cross-connected
> > >> cable between servers on eth1. I have both heartbeat bcast and drbd
> > >> working over eth1.
> > >>
> > >> I've had this working for a while and I've been testing various
> > >> fail-over
> > >> situations.
> > >>
> > >> The one thing I've noticed is that pretty much all hell breaks loose if
> > >> I
> > >> take down eth1 on either of the servers. While I would admit in a real
> > >> situation it is unlikely that anything should happen here, it can be a
> > >> point of failure.
> > >>
> > >> What are the best ways of address this? Is my setup lacking?
> > >>
> > >> _______________________________________________
> > >> drbd-user mailing list
> > >> drbd-user at lists.linbit.com
> > >> http://lists.linbit.com/mailman/listinfo/drbd-user
> > >
> > > _______________________________________________
> > > drbd-user mailing list
> > > drbd-user at lists.linbit.com
> > > http://lists.linbit.com/mailman/listinfo/drbd-user
> > >
> > _______________________________________________
> > drbd-user mailing list
> > drbd-user at lists.linbit.com
> > http://lists.linbit.com/mailman/listinfo/drbd-user
> drbd-user mailing list
> drbd-user at lists.linbit.com
More information about the drbd-user