I would simply stop all resource groups where they are, then power
down all the servers.  This way when the servers come up you can
easily restart the resource groups on thier normal node.

Denny

I would play it safe.  I am assuming that you are a running a multiple
instance failover configuration with each node hosting one of the SQL Server
instances.  With that being said, you probably want the SQL Server instances
to stay online and service client requests during the server reboots.  
Therefore, manually failover SQL Server Instance 01 from the first node to
the second node, and then reboot the first node.  Once node 1 is back online,
fail the SQL Server services back to node 1.  Repeat the steps for all nodes
within the cluster
I believe this rolling strategy works better than shutting down all of your
servers, SQL Server resources.  Or more than one node at the same time.  If
you shut down more than one node at the same time, It is highly probable that
surviving node would not have enough resources to sustain all of the failed
SQL Server instances without suffering performance degradation.  For example,
if you shut down node 1, 2 and 3 at the same time, most likely 4 would have
issues running all 4 SQL Server instances.

I hope that helps!!!

If you are going to take everyting offline for the reboot, then go ahead and
stop the resource groups and reboot everything.  If you want a rolling
reboot then move the SQl instances around to get one node empty, reboot it,
and repeat until all nodes are done.  End up by putting the instances back
on their preferred nodes.

Personally, I prefer to do a rolling reboot.  I have had cluster nodes
"stick" during boot when multiple nodes enumerated their connected storage
systems at the same time.