[DRBD-user] bonding driver
bthomas at wolfgeek.net
Thu Feb 5 18:37:01 CET 2004
For what it's worth, I'm running DRBD with bonding a pair of copper
gig-e connections. I can get a full sync on a quiesced system (no NFS
traffic, etc.) in about 12 hours total. Failover in heartbeat works as
I've not had any weirdness with heartbeat's behavior, but Phlipp
(ahem. :) ) is right -- What you're experiencing has nothing to do
with DRBD, you might want to poke around the heartbeat lists.
You say failback doesn't work properly, do you maybe have
nice_failback set to on in your ha.cf? If you do, this is the behavior
you'd see; since that's what nice_failback does. :)
On Thu, Feb 05, 2004 at 10:25:11AM +0100, Philipp Reisner wrote:
> Am Mittwoch, 4. Februar 2004 21:45 schrieb Ward Horner:
> > Has anyone ever used DRBD with the linux Bonding driver?
> > (/usr/src/linux-2.4/Documentation/networking/bonding.txt).I'm getting
> > somewhat strange results. For some reason, the primary node comes up as a
> > secondary and the secondary node comes up as the primary (ie the secondary
> > node serves the DRBD disk).
> > Normally when a primary node fails and the secondary node takes over, when
> > the primary node comes back on-line, it automatically takes back control
> > from the secondary node. This automatic take-over does not occur when using
> > the bonding driver. Fail over to the backup server works fine, but it never
> > changes back when the failed node comes back on-line.
> > When I remove the bonding driver from the system, then DRBD again works as
> > expected.
> Failovers/failbacks/switchovers are heatbeat's business.
> : Dipl-Ing Philipp Reisner Tel +43-1-8178292-50 :
> : LINBIT Information Technologies GmbH Fax +43-1-8178292-82 :
> : Sch?nbrunnerstr 244, 1120 Vienna, Austria http://www.linbit.com :
> drbd-user mailing list
> drbd-user at lists.linbit.com
More information about the drbd-user