Adding two standby APIC to a cluster of 5 nodes

[Bassam Abbasi]

Hello , 

 I am looking to expand a cluster of 3 controllers into 5 for leaf switches scalability in a stretch fabric setup, the 5 controllers will be split into two DCs (DC-A with 3 APICs & DC-B with 2) , however the shard distribution is a concern if DC-A  is down, there would be a possibility to lose a complete shard instances which is replicated on that DC. 

Thought of adding a standby controller in DC-B for one of the APICs in DC-A, but as far as I read there is no information replicated to the standby controller in advance,

"No information is replicated to standby controllers, including admin credentials." 

https://www.cisco.com/c/en/us/td/docs/switches/datacenter/aci/apic/sw/3-x/getting_started/b_APIC_Getting_Started_Guide_Rel_3_x/b_APIC_Getting_Started_Guide_Rel_3_x_chapter_0101.html 

so if DC-A is down and I bring up the standby controller in DC-B I will still not having those shards in DC-A .

Any suggestion to have all the shards available in DC-B upon a failure in DC-A? 

Best Regards. 

[gkumark]

Why don't you split your active controllers in to 2 sites and then have 1 standby controller in each site? That way the shards are available anytime at both the sites. In case of any site going down, you can just bring up the standby controller to the cluster?

-GK

[Bassam Abbasi]

Actually that is the plan, but since I will have 3 controllers in one DC, there is a chance that the 3 copies of certain shards will only exist on those three controllers. 

The concern if that DC is down we will lose those shards (no copies in the other DC), even If I have a standby controller in the second DC for an active controller in the first DC that won't solve the problem because there is no active syncing of shards until you activate the standby controller.