Hey, Peter...long time.
So, here goes...
I labbed this up about a month ago and have had so many things on my plate that I havent had a chance to play with it further. As an FYI, I dont do the day-to-day network engineering stuff anymore. I am a sales engineer and I work for a data center solutions company. Most of my life involves working with exciting, bleeding edge technologies, not this STP nonsense, which should disappear once TRILL is fully evolved. Anyway, I also dont remember everything I did. I will do it again either today or tomorrow and run my show commands again. If I recall right, the output of the show spanning tree command on the Cisco's indicated RSTP, not STP.
But I, too, was/am under the impression that the open standard version of RSTP has similar functionality (UF and BBF) either built in or that must be configured, but I dont see any counterpart to UF and BBF that I can configure, so if it exists, it must be built-in to the code. But if it does indeed exist, why does the port have to go into the LIS and LRN states? The port that does it is the DP on the RB.
I did document my lab set up for future reference. Let me post it here.
Hardware:
2 Cisco 3550s at the distro layer (D1 and D2), running 12.2(35)SE5.
1 Dell 6220 access switch (A1), running 2.2.0.3.
Topology:
6220 dual homed to each 3550
3550s connected to each other
All connections are dot1q trunks allowing all vlans (only VLAN 10 configured)
Cisco config includes:
Vlan 10
SVIs for Vlan 10
D1 is root for Vlan 10
D2 is Secondary for Vlan 10
Default RSTP timers
Rapid-pvstp+
Dot1q trunk downlinks to Dell 6220
Dot1q trunk between D1 and D2 (deliberately installed loop)
Dell config includes:
Vlan 10
rstp
dot1q uplinks (switchport mode general)
default RSTP timers
Converged STP status:
A1 to D1 link FWD
A1 to D2 link BLK (Cisco port blocked, Dell port in ALT)
D1 to D2 link FWD
D1 root bridge
D2's root port facing D1
A1's root port facing D1
Test 1:
Started continuous PING from A1 to SVI IP of D1
Disconnected A1 to D1 link --> failover to D2 link pretty fast (dropped one PING)
Reconnect A1 to D1 link ---> over 30 seconds to reconverge (about 18 PINGs dropped with a 2 sec timeout/PING)
My off-the-cuff suspicion:
Cisco’s rstp includes uplinkfast and backbonefast, Dell does not offer them, of course. So, there is no “agreement” created between the 3550 and the 6224. So, upon reconnecting the A1 to D1 link, the 3550 port facing the 6220 has to transition to LIS and LRN and then FWD – 2xfwd delay intervals.
So, I configured D1’s 6220-facing port for portfast (spanning-tree portfast trunk) and, of course, it transitioned immediately into FWD state and dropped only one PING when I reconnected the link. Not exactly a good practice to keep an inter-switch link set for portfast. :-)
Test 2:
Started continuous PING from A1 to SVI IP of D1
Installed same failure, but all switches now configured for 802.1D
Failover, as expected, was 50 seconds upon removal of A1 to D1 link and 50 seconds upon reconnection of A1 to D1 link.
Test 3:
Started continuous PING from A1 to SVI IP of D1
Kept 3550s in 802.1D (STP)
Disabled STP altogether from 6220 uplinks to 3550s.
Failover was 50 seconds upon removal of A1 to D1 link and 50 seconds upon reconnection of A1 to D1 link.
Debug feature not available on 6220s.
Debug on Ciscos reveal bi-directional BPDU exchanges with 6220 during reconvergence. Content not informative enough to draw conclusions.
