I assume that you are running more than 1 instance to be running active \ active.
In that case patch the passive instance on each cluster node first, then failover to the passive node. This will then perfom the updates to the system databases, so it will take a few minutes for the server to be fully operational again.
At that point patch the former active instances (now passive) and failback to ensure all patching has been successful.
Alternatively run both instances from one node of your cluster, patch both instances on the passive node, failover, then patch both nodes on the former active node, failover again to sensure all patching has been successful.
MCITP SQL 2005, MCSA SQL 2012