Replacing APIC whith Standby device

[waschminator]

Hello,

i am to replace an APIC controller by a standby device. after replacing it the former active device is powered down and the former standby apic is now active with the ID of the replaced device. so far so good.

next step is to wipe the old apic (acidiag touch clean/setup, acidiag reboot). my question is related to how to perform this step in detail: in the lab i simply booted up the old APIC and did the commands. but this means that the APIC gets connected to the fabric. in theory now 2 APIC-nodes with the same ID are connected to the fabric. officially this is a high risk because the cluster could get issues. in lab nothing happened, the old controller does not join the cluster. to be on the official safe side i would have to go to the datacenter and plug the cables...(or disable the interfaces in ACI). 

so the question is if it is really a risk to boot up the old APIC and wipe it(even if nothing happens) or if the new version now already have a mechanism implemented that would prohibit the old controller to join the fabric in case it was replaced by standby at the moment where the old controller was online.

for sure i understand in case of an unplanned outage i need to be carefull becuase then the failed controller does not know that it was replaced.

my current software version is 5.2(5c)

[tawa-ndafa]

Hello,

When you replace an active APIC with a standby one, the replaced APIC switches off but does not wipe its configurations, as you previously explained. I have performed this procedure a couple of times in production. While switching on the replaced APIC does not cause any issues in the production environment, I prefer to disconnect the cables when I am in a data center or shut down the interface on the leaf for safety reasons. Currently, there is no built-in system to automatically clear the configurations and prevent two APICs from coming up with the same configs.

I also doubt it will cause issues in prod because your cluster will 3 for example and there is no way a 4th APIC can come in.

[waschminator]

thanks for your feedback. cisco told us that with version 4 it was high risk because the database could be corrupted. i never used version 4 therfore i do not know. and there is not much information available public that provides information under the hood. 

for safety reasons we will disconnect the cables because we need to move APICs from one pod to another and will use the standby procedure for it.