https://balakumarnair.wordpress.com/2011/02/06/crsctl-start-crs-vs-start-cluster-2/

Starting Clusterware

In 11gR2, you can start the High Availability Service Daemon (OHASD) and Clusterware stack in a single step (A) or start it in a two step process(B).

(A) To start the OHASD & clusterware stack in a single step, use

$ crsctl start crs

Note that you can start components on the local server only with this.

(B) You can do this in two steps; starting OHASD and then starting Clusterware stack. The advantage of using this method is that, if OHASD is running, you can start the clusterware stack on any node from local node.

You can start OHASD by,

$ crsctl start has

then, you can start the clusterware stack by,

$ crsctl start cluster -all

Syntax

crsctl start cluster [-all | -n server_name [...]]

-all Start clusterware on all nodes
-n Start clusterware on particular nodes

If -all or -n is note used, then the clusterware is started only on the local node.

Stopping Clusterware

In 11gR2, you can stop the High Availability Service Daemon (OHASD) and Clusterware stack in a single step (A) or stop it in a two step process(B).

(A) To stop the OHASD & clusterware stack in a single step, use
$ crsctl stop crs

If any of the process managed by oracle Clusterware is still running after executing it, then the command fails. Use the force option in such cases to terminate them unconditionally.

$ crsctl stop crs –all –f

Note that you can stop components on the local server only with this.

(B) You can stop processes in two steps; stoping OHASD and then stoping Clusterware stack. The advantage of using this method is that, you can stop the clusterware stack on any node from local node.

You can stop the clusterware stack by,

$ crsctl stop cluster -all

Syntax

crsctl stop cluster [-all | -n server_name [...]] [-f]

-all Stop clusterware on all nodes
-n Stop clusterware on particular nodes
-f Use this to stop clusterware resources unconditionally if 'crsctl stop cluster' fails to stop.

If -all or -n is note used, then the clusterware is stopped only on the local node.

Once the clusterware stack is down then you can stop OHASD by,

$ crsctl stop has

The advantage of crsctl stop cluster compared to stop crs command is that it prevent the relocation of 'certain' resources to other servers in the cluster before the Clusterware is stopped on that server.

Status

[root@rac2 ~]# crsctl status server -f
NAME=rac1
STATE=ONLINE
ACTIVE_POOLS=Generic ora.RACDB
STATE_DETAILS=

NAME=rac2
STATE=ONLINE
ACTIVE_POOLS=Generic ora.RACDB
STATE_DETAILS=AUTOSTARTING RESOURCES