Hi,
We run a private datacenter with EX4500 VC at the core and EX4200 VC at the edges.
On the EX4200 stacks, we've got hundreds of vmware/Dell servers with multiple 1G attachments.
We don't use LACP on server attachments (load distribution is achieved by vmware/vsphere proprietary algorithms, distributing vm mac adresses on network uplink). It's been running for years without any problems.
We usually don't accept on our network other hosts/devices than those we test and install by ourselves.
For the first time we are asked to offer network connection to some custumer exotic devices... and they want 4x1G LACP attachements on the EX4200 edge switches.
Ten years ago, I had a very bad experience in such a configuration when a customer plugged its own (nortel) switch with multiple attachements into our core campus switch.
The Core was configured with static aggregation (no LACP, old style MLT aggregation), the customer misconfigured its switch and forgot to validate link aggregation on his side... boom... loop... campus down.
Today, as we will configure the AE interface with LACP (active mode) I guess this couldn't happen today with LACP, even if the customer misconfigured its devices and forgot to validate link aggregation or LACP.
I'd like to discuss how an EX4200 with LACP would act in such a situation ?
=> Our side, EX4200 4x1G, LACP VS customer, misconfigured, 4x1G, no link aggregation, no LACP :
LACP shouldn't come up on our side.
=> If the customer is transmitting traffic/broadcasts over its misconfigured non agregated ports, what will happen ? Will the EX4200 drop these packets (because they're being received on interfaces part of a LAG that is down, even if each individual interface is up) or accept them and switch them on individual interfaces, hence creating a loop ?
ence creating a loop ?
In other words, does LACP prevent loop formation in such (mis)configuration ?
Regards,
Pascal