Adding +2 switches N9K-C9364C to a SPINE layer

[ss1]

Dear Friends,

My network is entirely based on VxLAN bridging and the current spine layer consists of two N9K-C9364C. However as the network grown our traffic exceeded 2T through these switches combined, hence I'm looking to add an additional set of two switches to the said topology. The spine layer only does PIM and OSPF routing towards 80 switches N9K-C9396PX, N9K-C9372PX and N9K-C93180YC-EX so it doesn't do anything else like VTEPs or something. Just pure routing and BGP route reflector sessions to all the leafs used as ToRs.

There is an anycast-rp configuration used on both spines right now:
ip pim rp-address 10.128.1.1 group-list 225.0.0.0/8 override
ip pim anycast-rp 10.128.1.1 10.128.2.183
ip pim anycast-rp 10.128.1.1 10.128.2.184

So basically what I'm looking to do is to add another 2 switches to the rp-set, let's say:
ip pim anycast-rp 10.128.1.1 10.128.2.153
ip pim anycast-rp 10.128.1.1 10.128.2.154
in order to have all four switches to the rp-set.

Then I will plan to enable the OSPF to the new spine switches and the links to the leafs as well. Just in case I will have ports between all spines in a mesh because some switches don't have the capabilities to be linked to all 4 spines at the same time. The new switches will be pre-configured with the same anycast-rp config done on the existing.

At the time I add the new spines to the RP-SET to the existing switches, will the existing switches do nothing because the new spines are not yet alive in the topology? Then, what will be the expected result once I enable the links to the new spines and they converge into the topology? Is there some danger of services outage or traffic blackholing especially if some leafs do not have links to the new spines? How risky may this operation be? 

Thank you all for sharing some experience and best practices on this!

[00u1661g7uPHO4xrg5d7]

Are you using any tools to manage the Fabric, If not I would recommend using Cisco NDFC tool which is very handy and easy to manage Cisco Nexus fabric operations and additions.

Also instead of adding Spine in same fabric u can try to use Multisite Fabric if that option is available to you. 

If you only have option to add in existing fiber then schedule during maintained window to avoid any outages.

[ss1]

Sadly we are not using any tools to manage them apart from standard CLI.
Do you think my proposed workflow procedure is the correct one? 

[ss1]

Dear Friends,

Do you think my approach on how to add new spines to the anycast-rp set will be correct?

Thanks
Stefan