Hi all, We have the a proposed design of 9x routers in a ring topology.  These routers would be in 9x different data centers spanning a geographical area the size of a large USA state. The design is the same as the design at this link, but instead of 4x routers there are 9x in the ring topology: https://learningnetwork.cisco.com/docs/DOC-11844 We were wondering if there was a better way to do iBGP with these 9x routers so we could cut down on the number of iBGP neighbor relationships that would be required for a full-mesh.  This topology could grow to 18x total routers in the future when an additional router is added to each site for resiliency.  That would be a lot of peering for a full-mesh. A iBGP design with route-reflectors doesn't seem to make much sense with a ring topology.  We mocked up a BGP confederation design where each site was its own sub-AS and it only peered with its directly connected neighbor.  That seemed like a good idea and would scale very nicely when the number of routers doubled to 18x.  At that point we would have two routers in each sub-AS. However, as nicely pointed out in the link above (https://learningnetwork.cisco.com/docs/DOC-11844), the path selection for BGP was sub-optimal.  If if a route was one hop away, it sometimes would traverse the long way around the ring as the "best" path.  That is because sub-AS's don't count towards the AS Path length.  This is very well explained in the link, so i won't explain further. Is there a better way to design this or are we just stuck with a full mesh of possibly 18x or more routers?  Is doing eBGP between sites a better idea?  Two of the sites would however need public ASNs as they will have Internet connectivity for the ring. Thanks! Sean  ... View more

We have a Nexus 7700 running two different contexts.  One is the Admin context and then another standard context.  They are configured with IP addresses within the same subnet (10.101.0.0/24), but they can't communicate with each - other ping doesn't even work between them.   However, they can both talk to all other devices on that subnet just not each other...  I've included the output from some commands below along with the NX-OS version.  Both mgmt0 interfaces in each VDC have different MAC addresses as they should. Is this a bug or am I missing something here?  The upstream switchport is just an access port. Thanks for any help/insight anyone can provide! Sean   version 6.2(8a)   Admin Context show run int mgmt0 interface mgmt0   vrf member management   ip address 10.101.0.10/24 show int mgmt0 mgmt0 is up admin state is up   Hardware: GigabitEthernet, address: 8480.2da5.f3c4 (bia 8480.2da5.f3c4)   Internet Address is 10.101.0.10/24   MTU 1500 bytes, BW 1000000 Kbit, DLY 10 usec   reliability 255/255, txload 1/255, rxload 1/255   Encapsulation ARPA, medium is broadcast   Port mode is routed   full-duplex, 1000 Mb/s   Auto-Negotiation is turned on   Auto-mdix is turned off   EtherType is 0x0000   1 minute input rate 32 bits/sec, 0 packets/sec   1 minute output rate 24 bits/sec, 0 packets/sec   Rx     188445 input packets 6622 unicast packets 181375 multicast packets     448 broadcast packets 48204530 bytes   Tx     90779 output packets 20 unicast packets 90737 multicast packets     22 broadcast packets 21779795 bytes       Other Context show run int mgmt0 interface mgmt0   vrf member management   ip address 10.101.0.11/24 show int mgmt0 mgmt0 is up admin state is up   Hardware: GigabitEthernet, address: 8480.2da5.f3c5 (bia 8480.2da5.f3c5)   Internet Address is 10.101.0.11/24   MTU 1500 bytes, BW 1000000 Kbit, DLY 10 usec   reliability 255/255, txload 1/255, rxload 1/255   Encapsulation ARPA, medium is broadcast   Port mode is routed   full-duplex, 1000 Mb/s   Auto-Negotiation is turned on   Auto-mdix is turned off   EtherType is 0x0000   1 minute input rate 696 bits/sec, 0 packets/sec   1 minute output rate 696 bits/sec, 0 packets/sec   Rx     5647969 input packets 5466145 unicast packets 181376 multicast packets     448 broadcast packets 495885390 bytes   Tx     5569760 output packets 5466115 unicast packets 92181 multicast packets     11464 broadcast packets 474893154 bytes     ... View more

We have an issue where switches are failing weekly in a switch closet.  In the past month we have gone through several 3750G switches and a couple 4510s.  The power supplies have eventually made a popping noise and had to be replaced.  on the 4510s we've tried two chassis and gone through several power supplies. The switches have been behind UPS systems so should be receiving conditioned power. Could load from the PoE devices really be causing this?  I wouldn't think it's power since they are behind a UPS. Any ideas on how to tackle this issue?  We're going through switches like crazy... Thanks Sean ... View more