[ovs-dev] [PATCH] bond: Only drop packets that indicate moves on SLB bonds.
ethan at nicira.com
Fri Nov 4 12:04:05 PDT 2011
> Yes, that makes sense to me. I thought that we had essentially agreed
> on that change earlier in fact.
Oh I'd forgotten about that. I'll go ahead and create a bug for the backlog.
> On Fri, Nov 04, 2011 at 12:00:26PM -0700, Ethan Jackson wrote:
>> This looks good to me.
>> Unrelated to this patch:
>> The more I think about it. I don't think falling back to balance-slb
>> is the appropriate thing to do when lacp negotiations fail for
>> balance-tcp bonds. I think it will be much safer to fall back to
>> Generally speaking, in a properly configured system, LACP negotiations
>> won't fail. Therefore LACP negotiation failures represent exceptional
>> circumstances in which safety seems like it would extremely valuable.
>> I'm worried about the case where someone configures their network with
>> a LACP bond going to 2 or more separate switches (completely valid
>> according to the spec). If something goes wrong, they will fall back
>> to balance-slb and thus be running an slb bond in a distributed manner
>> across multiple switches. This should "theoretically" work, but feels
>> risky to me. I'm confident active-backup will work in all cases
>> On Fri, Nov 4, 2011 at 09:51, Ben Pfaff <blp at nicira.com> wrote:
>> > On Fri, Oct 28, 2011 at 01:28:57PM -0700, Ben Pfaff wrote:
>> >> SLB bonds, for important reasons, drop most incoming packets that indicate
>> >> that a MAC has moved to the bond from another port. ?These reasons do not
>> >> apply to other types of bonds, but until now OVS has still dropped them.
>> >> This fixes the problem. ?It changes behavior of active-backup bonds and
>> >> stable bonds, neither of which has the same problem as SLB. ?Behavior of
>> >> SLB bonds and TCP bonds in SLB fallback mode is unaffected.
>> >> Bug #7928.
>> > This needs review. ?It shouldn't be very hard.
More information about the dev