The first step of
the migration is to upgrade all spine switch software.
Before You Begin The following pre-requisites
must be met for upgrading Nexus 6000 spine switch software
Nexus 6000
series switch must be running on Cisco NX-OS version 6.2.(2) software release
Border leaf
nodes must be DFA hardware-capable Nexus 6000 nodes running Cisco NX-OS version
6.0.2.N2
Note
If you have
anything other than a Nexus 6000 series switch, you must physically replace the
switch with the version 7.0.(0)N1(1) image; the configuration remains the same
as the previous image.
Step 1
On the Nexus
6000 series spine switches, perform a non-disruptive in-service software
upgrade (ISSU) upgrade to Cisco NX-OS version 7.0(0)N1(1).
Make sure you have
upgraded the border leaf software, as described in Step 2.
Step 1
On the first
border leaf switch:
Configure an
additional hot standby router protocol (HSRP) per virtual local area network
(VLAN) with the anycast gateway MAC address with an unused IP adddress.
Configure
the iBGP router reflector
Configure
anycast gateway MAC
Add a
vrf-tenant-profile and configure the virtual network identifier (VNI) under the
virtual router
Enable
traditional forwarding on the switch virtual interfaces (SVIs)
Advertise
host routes to the BGP route reflector
Step 2
Repeat step 1
a-f for the second border leaf switch in the pair.
There should be no
change to the traffic flow after you have configured the border leaf pair.
What to Do Next
Upgrade the software
for the second pair of border leaves.
Step 4: Upgrade the
Second Border Leaf Pair
In this procedure,
you will perform a disruptive in-service software upgrade (ISSU) for the second
border leaf pair.
Make sure you have
upgraded the border leaf software, as described in Step 4.
Step 1
On the first
border leaf switch in the pair:
Configure an
additional hot standby router protocol (HSRP) per virtual local area network
(VLAN) with the anycast gateway MAC address with an unused IP address.
Configure
the iBGP router reflector
Configure
the anycast-gateway MAC address
Add a
vrf-tenant-profile and configure the virtual network identifier (VNI) under the
virtual router
Create
switch virtual interfaces (SVIs), if they are not present.
Enable
anycast-gateway on the SVIs.
Enable
traditional forwarding on the switch virtual interfaces (SVIs)
Advertise
host routes to the BGP route reflector
Step 2
Repeat step 1
a-h for the second border leaf switch in the pair.
After you configure
the second border leaf pair, the following changes occur:
All border leafs
start to respond to address resolution protocol (ARP) with anycast-gateway MAC
addresses.
Direct
forwarding occurs between the border leaf pairs, without going through the
Nexus 7000.
The unicast traffic
flow and North-South traffic remains unchanged.
What to Do Next
Upgrade the software
for FabricPath leaf pair.
Step 6: Upgrade the
FabricPath Leaf Pair
In this procedure,
you will perform an in-service software upgrade (ISSU) for the FabricPath leaf
pair.
Prior to configuring
the Fabric Path leaf, you should have upgraded the software.
Step 1
On the first
switch in the pair:
Install a L3
license.
Enable Cisco
DFA.
Add the iBGP
router reflector client.
Add the
segment ID and VRF.
Add a
vrf-tenant-profile and configure the virtual network identifier (VNI) under the
virtual router.
Create
switch virtual interfaces (SVIs) for all VLANs.
Enable
anycast-gateway for all VLANs.
Advertise
host routes to the BGP route reflector.
Start
handling gateway functionality for local hosts.
Step 2
Repeat step 1
a-i for the second border leaf switch in the pair.
If you are migrating
a fabric that includes both Nexus 5000 and Nexus 6000 switches:
Migration is
completed if you have upgraded all Nexus 6000 software and enabled Cisco DFA
forwarding.
HSRP/VRRP will
remain as long as there are Nexus 5000 leaves in the network.
In Nexus
5000-involved VLANs and SVIs: VLANs are global, non-segment-id-enabled, and the
forwarding mode can be either proxy or anycast gateway.
In upgraded and
configured Nexus 6000-involved VLANs and SVIs: Can be segment-id enabled, and
the forwarding mode can be either proxy or anycast gateway.
Multicast will
continue run in the legacy multicast mode. Cisco DFA multicast should not be
enabled.
What to Do Next
If you are migrating
a fabric that includes both Nexus 5000 and Nexus 6000 switches, the migration
is completed if you have upgraded all Nexus 6000 software and enabled Cisco DFA
forwarding.
Step 8: Upgrade and
Configure All Remaining Leaf Switches
You should perform
this procedure on all of the remaining leaf switches in the network. Refer to
Steps 6 and 7 for additional information.
Upgrade the
software on all of the remaining leaf switches.
Step 2
Add Cisco
DFA-related configuration on all of the remaining leaf switches.
Step 3
Enable
anycast-gateway on leaf switches for all VLANs.
What to Do Next
Remove HSRP
Configurations on border leaf pairs.
Step 9: Remove HSRP
Configurations on Border Leaf Pairs
During the
migration, some hosts learn the anycast gateway MAC address as its MAC address
for the default gateway. Some hosts will learn the HSRP VMAC as the MAC for the
default gateway. We recommend that you wait a couple of hours to make sure that
the HSRP VMAC is aged out on all hosts.
You should have
completed migration on all leaf switches.
Step 1
Remove HSRP
configuration on each border leaf switch.
Step 2
Change the SVI
IP address to the VIP.
After you have
removed HSRP configurations, migration is complete.
You can move to
Cisco DFA multicast, if preferred. There is no VPC on border leaf connector to
external multicast routers.
SVIs can be
switched to proxy forwarding mode, if preferred.
New VLANs can be
segment-id enabled.
In the all-Nexus
6000 topology, you can move to Cisco DFA multicast, if preferred.
Note
Multicast
traffic disruption will occur during the move to Cisco DFA multicast. Also, the
border leaf that connects to external multicast routers cannot run VPC.