Start a Conversation

Unsolved

This post is more than 5 years old

O

8372

October 11th, 2012 23:00

Implemeting 802.3ad on 5324

Hi,

i have 2 5324 interconnected with 1 trunk port. Now i have 1 linux box on which i created 802.3ad bond and bridge. Each cable of the linux bond is connected to a different switch and nothing special is configured on those ports. With that config i am getting broadcast storms.

My question is, can i configure LACL on only 1 port per each switch and expect that this works or i have to configure 2 ports on each switch with LAC and join LAG so the bond cables from linux machine enter ONLY 1 switch and for 2nd switch i would need 2 additionaly bonded nics?

 



"With modes balance-rr, balance-xor, broadcast and 802.3ad all physical ports in the link aggregation group must reside on the same logical switch, which in most scenarios will leave a single point of failure when the physical switch to which both links are connected goes offline. Modes active-backup, balance-tlb, and balance-alb can also be set up with two or more switches"
"Bear in mind that with 802.3ad all of the links in your bond need to be connected to the same switch, unless the switch supports sharing the 802.3ad information with other switches through a stack/virtual switch type configuration"
 
 
Can LAG span across 2 or more Dell 5324 switches?

18 Posts

October 12th, 2012 09:00

my apologies in advance if i have not understood this correctly,

but ..

for LAG/channel groups, you have to ensure that you are uplinking to the same switch (rather, the same logical chassis) upstream.

thus, your uplinks would need to go to the same upstream switch.

UNLESS..

you have the two upstream switches working as a stack (Juniper Lingo Virtual Chassis, Cisco Lingo VSS/Stackwise, PC lingo stacking). In this situation, they are a single logical fabric, despite being two different physical switches. thus, you would be able to build the LAG/channel across the two.

now, stacking is not the same as trunking. when you interconnect the switches via trunking, they are still 2 distinct logical chassis. unless you utilize stacking (if the feature is available on the model), they would remain distinct. some switches implement stacking through front ports (PowerConnect 80xx), some through rear dedicated modules (62xx). i dont remember how 53xx does it off the top of my head, but cli & getting started guides for the model should clarify this.

the last variation on your model is VLT/vPC. you can have two distinct upstream switches and still form a LAG from a downstream client to the two switches, if they support what Cisco calls and implements as vPC, and Force10 calls a VLT. the beauty of vPC/VLT is that the two switches remain distinct logical switches from the perspective of their upstream device, but for their downstream devices, they appear as one logical chassis.

i hope this helps, but as i said, if i have misunderstood your query, then my apologies.

28 Posts

October 12th, 2012 11:00

Hi ricther, thanks for your answer.

This is my 1st time i am trying to implement this so i have a lot of questions on this matter.

The whole picture is that i actually  have 3 switch pairs (each switch pair is connected with a trunk link but it's not connected with trunk link to other switch pairs) and i have 4 linux servers with 6 NICs on each server. 2 NICs are bonded and bridged so, each bond is connected to every switch pair (1 cable to 1 switch, other cable to different switch of the same switch pair which is not proper connecting looks like). It's a bit complex architecture but what i trying to achieve here is bond of 2Gbit and redundancy.

My next question would be since i can't have stacking on this switch, i need to configure 2 ports on the SAME switch for LACP and create LAG correct?. Linux bond must be plugged into those LAG ports correct? Now since 2 more bonded NICs will be connected to that same switch, can i add 4 more ports to same LAG group and just plug those bonded nics into those LAG port groups? Lastly probably i MUST enable STP on those bonds on each Linux machine since without it would mean switching loops all the time ?

No Events found!

Top