General Guidelines
Before configuring the inter-rack redundancy deployment, here are some general guidelines:
-
Both racks should be on the same software version.
-
Both racks should be configured with same configuration.
Loopback port of Instance 1 and Instance 2 should be different. Else, REST-EP POD would not come up due to K8 IP/Port conflict.
-
Respective interface on both the racks should be on the same VLAN. For example, N4 VLAN of Instance1 and Instance2 should be on the same VLAN. Else, there is a route conflict on Kernel while enforcing BGP policies.
-
Consult your Cisco Technical Representative to perform the following procedures to make sure proper roles are assigned.
For more information, see Software Upgrade on GR Pairs.
-
Post failover, perform the failback manually after ensuring the rack is healthy. Autonomous failback is not supported.
For more information, see Recovery Procedure.
-
Use non-bonded interface in BGP speaker PODs for BGP peering.
-
BGP peering per Proto node is supported with only two BGP routers/leafs. Considering two Proto nodes, there can be maximum of four BGP neighborships.
-
Use bonded interfaces for Service traffic.
-
Geo pod uses two VIPs:
-
Internal-VIP for Inter-POD communication (within the rack)
-
External-VIP for Inter-rack Geo pod communication. Configure only on Proto Nodes on L2 Subnet. This is used to communicate across the racks. This node has external connectivity to other rack.
-
-
Geo Internal IP to be reachable to all nodes within the rack.
-
Geo External IP:
-
CDL/Kafka VIPs: Configure on CDL Labeled Nodes on L2 Subnet.
-
Enable LI tapping on both the racks.
-
MDF server should be reachable from both the racks.