[c-nsp] Dell switches (specifically PowerConnect 7048P) and Ciscos

[Phil Mayers]

On 11/27/2012 03:40 AM, Kell, Jeff wrote:
> We're doing an eval on some PowerConnect 7048P switches, and have run
> into spanning tree issues.  They don't like PVST, but will spit out
> STP that "in theory" will revert a Cisco to STP (is this process
> contagious?  or limited to the upstream?).

The STP process on vlan1 emits both Cisco PVST and IEEE PDUs IIRC, and 
when it detects a PDU from the other end on a given port, switches to 
just using that version of the protocol until the STP state machine is 
reset (i.e. link down). I think there's also a CLI command to blip it.

So it's a per-port (well, per virtual-port, really) thing, similar to 
the PVST/MST compat. mode that Ciscos do when running MST.

>
> That unfortunately requires an untagged vlan 1 on all trunk ports
> (breaks our standards and would require a mass reconfiguration).

Yeah, I always thought it was a bit sucky that the PVST fallback didn't 
take place on the untagged/native VLAN per-port, as opposed to being 
hardcoded to vlan1. How hard can it be to implement that?


However: IIRC the IEEE-fallback is of limited use in some topologies. If 
the downstream (IEEE-only) switch blocks the port, then it blocks it for 
all vlans. But, those vlan/port combos on the upstream Ciscos are in FWD 
because they can't "hear" each other via that path. This can lead to 
surprising behaviour.

Worse, if the blocked port is on the upstream Cisco, it's only blocked 
for vlan 1 using this method. The other vlans are blocked by virtue of 
the Ciscos hearing each others PVST via that path, and this comes with 
the attendant 6/30 second delay on failover/failback.

>
> You can also try MST if you want to convert (again, breaks our
> standards and would require a mass reconfiguration).
>
> Otherwise they seem quite tolerable and have some attractive
> features...  but I'm not quite sold if it requires a campus-wide mass
> reconfiguration to get along with existing gear.

Personally I wouldn't go for it; but what the question? ;o)

Normally I'm not a big fan of "proprietary" protocols, but MST is so 
awesomely sucky for Campus environments ("map all your VLANs to 
instances before you start, and never change it" - yeah, right!) that we 
mandate Cisco compatible PVST in all our edge.