Question about 802.3ad bonding mode.

Question about 802.3ad bonding mode.

[Dmitry Krivenok]

Hello,
I have two physical NICs bonded into bond0 interface with bonding
mode=4 (i,e, 802.3ad).
Both NICs are connected to the same switch and switch ports are *not*
configured in "lacp" mode.

Documentation on this mode says this:

The 802.3ad mode requires that the switch have the appropriate
ports configured as an 802.3ad aggregation. The precise method used
to configure this varies from switch to switch, but, for example, a
Cisco 3550 series switch requires that the appropriate ports first be
grouped together in a single etherchannel instance, then that
etherchannel is set to mode "lacp" to enable 802.3ad (instead of
standard EtherChannel).

I'd expect that in such a case LAG should not be created via LACP. And
it's actually not created because I see the following LACPDUs coming
out of my NICs every 30 seconds:

13:27:25.968094 LACPv1, length: 110
        Actor Information TLV (0x01), length: 20
          System 00:60:16:56:ab:c1, System Priority 65535, Key 17,
Port 1, Port Priority 255
          State Flags [Activity, Aggregation, Synchronization, Default]
        Partner Information TLV (0x02), length: 20
          System 00:00:00:00:00:00, System Priority 65535, Key 1, Port
1, Port Priority 255
          State Flags [Activity]
        Collector Information TLV (0x03), length: 16
          Max Delay 0
        Terminator TLV (0x00), length: 0

AFAIK state flag "Default" means that we never got any LACPDUs from the switch.

I have IP address configured on bond0 and from what I see it's fully
working. I can ping it from various hosts, I can ssh to that host,
etc. From user perspective it works just fine except that there is no
LAG...

Is that correct behavior of dynamic link aggregation via LACP?
Shouldn't it be configuration error? Is it 802.1ax compliant?

Any help is very appreciated.

Thanks,
Dmitry

P.S.
My host is SLES11SP2, kernel 3.0.

Re: Question about 802.3ad bonding mode.

[Jay Vosburgh]

Dmitry Krivenok <krivenok.dmitry@gmail.com> wrote:

>Hello,
>I have two physical NICs bonded into bond0 interface with bonding
>mode=4 (i,e, 802.3ad).
>Both NICs are connected to the same switch and switch ports are *not*
>configured in "lacp" mode.
[...]
>I have IP address configured on bond0 and from what I see it's fully
>working. I can ping it from various hosts, I can ssh to that host,
>etc. From user perspective it works just fine except that there is no
>LAG...
>
>Is that correct behavior of dynamic link aggregation via LACP?
>Shouldn't it be configuration error? Is it 802.1ax compliant?

	Your system is behaving correctly; there is a backwards
compatibility mechanism in the standard.

	Essentially, if there is no LACPDU exchange, the affected links
are marked as "Individual" (as opposed to "Aggregatable"), which means
that those links become members of an aggregator with exactly one
member: that link.

	Then, when the active aggregator to use for traffic is selected,
if all links are Individual, then one of those is selected, and there is
then one link that works for communication with the partner system.

	This is done intentionally to permit communication from a system
configured for LACP to a system that is not so configured.

	This is mentioned at a high level in IEEE 802.1AX section 5.1.2
(j), the Individual vs Aggregatable is in 5.3.5 and 5.3.6, the treatment
of Individual links is in 5.3.9, and 5.4.12 says, in part:

	If no LACPDU is received before the current_while timer expires
	again, the state machine transits to the DEFAULTED state. [...]
	This allows configuration of aggregations and individual links
	when no protocol Partner is present, while still permitting an
	active Partner to override default settings.  [...]

	-J

---
	-Jay Vosburgh, jay.vosburgh@canonical.com