Hello,
You did not have to close the previous thread I just was on a business trip yesterday and did not have time to respond.
1.) Since the N5Ks are running vPC and form a virtual chassis, and BS1
and BS2 are stacked, what you have is a virtual chassis in each layer.
So, each layer sees the other as ONE switch. The N5Ks see the BS
switches as 1 virtual switch and the BS switches see the N5Ks as 1
virtual switch. Therefore, there is no loop and STP is only acting as a
preventive mechanism.
If I understand you correctly, the BS1+BS2 are stacked (behaving as a single switch), and N5K-1+N5K-2 form a virtual single chassis (also behaving as a single switch). Then, however, you have a "BS" switch and a "N5K" switch interconnected with two uplinks (BS1/N5K-1, BS2/N5K-2). I believe that is a loop and the STP is mandatory in such a case.
2.) Since BS1 and BS2 are running IEEE STP (which only has 1 STP
instance, namely, the CST) and the rest of the network is running
PVST+, D1 OR D2 must act as the root bridge for ALL VLANs, since the
IEEE STP only understands one logical STP topology and all VLANs are
mapped to it.
I believe that you can configure the BS switch to act as STP root with no ill effects. This switch will in turn become the root also for VLAN1 in the PVST+ region. Other switches in the PVST+ region will become roots for other individual VLANs. I do not think you have to configure D1 or D2 to act as root switches for all VLANs.
3.) Loopguard, UDLD aggressive, BPDUGUARD and any other Cisco
proprietary STP add-on will not be recognized by BS1 and BS2 and will
not participate in any conversations that those protocols engage in.
Loopguard does not send any special datagrams - it is simply about observing the arrival of BPDUs and taking certain countermeasures when they suddenly stop arriving. Thus, the is actually transparent to non-Cisco switches. BPGUGuard is at the same level - it reacts to arrival of an unexpected BPDU on a guarded port, nothing more. Again, this is transparent to non-Cisco switches. For UDLD to work, both switches on a point-to-point link must be configured for UDLD monitoring. It does not make to configure the UDLD only on a single side of a link - that would not provide any protection. Cisco UplinkFast or BackboneFast are proprietary extensions and would not be understood by non-Cisco switches (these conversations will be tunneled).
4.) An alternative to configuring either D1 or D2 as the root for all
VLANs, we can reconfigure the environment to run MSTP, since only 2
logical STP topologies exist anyway -- one in which D1 is the root and
the second in which D2 is the root. In the case of MSTP, two MSTI
instances will be created, in which half the VLANs may belong to one
and the other half to the other. In addition, one IST will be created
and that acts as the STP instance that communicates with the CST of the
IEEE STP.
As I indicated earlier, my feeling is that you do not need to have D1/D2 be the root for all VLANs. You can if you need, but you don't have to. Otherwise, you are correct in your assumptions here. However, you won't avoid the troubles with PVST+ when you migrate to MSTP only on D1+D2 - a Cisco switch running MSTP performs MSTP-to-PVST+ translation on boundary ports. If you'd like to migrate to MSTP, then do it completely - make all switches run MSTP, which is recommended anyway (combining different STP types in a single switched network is in my opinion not the way to go).
5.) A 3rd approach may be to create an MST region out of BS1, BS2, and
N5K-1 and N5K-2. In this case, D1 and D2 will see that MST region as
one logical CST bridge. This solution seems a bit clumsy.
No mixing of STP incarnations. You don't want to have your headaches multiplied by the count of STP versions running in your network
Best regards,
Peter
