11gR2 Clusterware and Grid Home - What You Need to Know

11gR2 Clusterware Key Facts

11gR2 Clusterware is required to be up and running prior to installing a 11gR2 Real Application Clusters database.
The GRID home consists of the Oracle Clusterware and ASM. ASM should not be in a separate home.
The 11gR2 Clusterware can be installed in "Standalone" mode for ASM and/or "Oracle Restart" single node support. This clusterware is a subset of the full clusterware described in this document.
The 11gR2 Clusterware can be run by itself or on top of vendor clusterware. See the certification matrix for certified combinations. Ref: Note: 184875.1 "How To Check The Certification Matrix for Real Application Clusters"
The GRID Home and the RAC/DB Home must be installed in different locations.
The 11gR2 Clusterware requires a shared OCR files and voting files. These can be stored on ASM or a cluster filesystem.
The OCR is backed up automatically every 4 hours to <GRID_HOME>/cdata/<clustername>/ and can be restored via ocrconfig.
The voting file is backed up into the OCR at every configuration change and can be restored via crsctl.
The 11gR2 Clusterware requires at least one private network for inter-node communication and at least one public network for external communication. Several virtual IPs need to be registered with DNS. This includes the node VIPs (one per node), SCAN VIPs (three). This can be done manually via your network administrator or optionally you could configure the "GNS" (Grid Naming Service) in the Oracle clusterware to handle this for you (note that GNS requires its own VIP).
A SCAN (Single Client Access Name) is provided to clients to connect to. For more information on SCAN see Note: 887522.1
The root.sh script at the end of the clusterware installation starts the clusterware stack. For information on troubleshooting root.sh issues see Note: 1053970.1
Only one set of clusterware daemons can be running per node.
On Unix, the clusterware stack is started via the init.ohasd script referenced in /etc/inittab with "respawn".
A node can be evicted (rebooted) if a node is deemed to be unhealthy. This is done so that the health of the entire cluster can be maintained. For more information on this see: Note: 1050693.1"Troubleshooting 11.2 Clusterware Node Evictions (Reboots)"
Either have vendor time synchronization software (like NTP) fully configured and running or have it not configured at all and let CTSS handle time synchronization. See Note: 1054006.1 for more information.
If installing DB homes for a lower version, you will need to pin the nodes in the clusterware or you will see ORA-29702 errors. See Note 946332.1 and Note:948456.1 for more information.
The clusterware stack can be started by either booting the machine, running "crsctl start crs" to start the clusterware stack, or by running "crsctl start cluster" to start the clusterware on all nodes. Note that crsctl is in the <GRID_HOME>/bin directory. Note that "crsctl start cluster" will only work if ohasd is running.
The clusterware stack can be stopped by either shutting down the machine, running "crsctl stop crs" to stop the clusterware stack, or by running "crsctl stop cluster" to stop the clusterware on all nodes. Note that crsctl is in the <GRID_HOME>/bin directory.
Killing clusterware daemons is not supported.
Instance is now part of .db resources in "crsctl stat res -t" output, there is no separate .inst resource for 11gR2 instance.

Note that it is also a good idea to follow the RAC Assurance best practices in Note: 810394.1

Clusterware Startup Sequence

The following is the Clusterware startup sequence (image from the "Oracle Clusterware Administration and Deployment Guide):

Don't let this picture scare you too much. You aren't responsible for managing all of these processes, that is the Clusterware's job!

Short summary of the startup sequence: INIT spawns init.ohasd (with respawn) which in turn starts the OHASD process (Oracle High Availability Services Daemon). This daemon spawns 4 processes.

Level 1: OHASD Spawns:

cssdagent - Agent responsible for spawning CSSD.
orarootagent - Agent responsible for managing all root owned ohasd resources.
oraagent - Agent responsible for managing all oracle owned ohasd resources.
cssdmonitor - Monitors CSSD and node health (along wth the cssdagent).

Level 2: OHASD rootagent spawns:

CRSD - Primary daemon responsible for managing cluster resources.
CTSSD - Cluster Time Synchronization Services Daemon
Diskmon
ACFS (ASM Cluster File System) Drivers

Level 2: OHASD oraagent spawns:

MDNSD - Used for DNS lookup
GIPCD - Used for inter-process and inter-node communication
GPNPD - Grid Plug & Play Profile Daemon
EVMD - Event Monitor Daemon
ASM - Resource for monitoring ASM instances

Level 3: CRSD spawns:

orarootagent - Agent responsible for managing all root owned crsd resources.
oraagent - Agent responsible for managing all oracle owned crsd resources.

Level 4: CRSD rootagent spawns:

Network resource - To monitor the public network
SCAN VIP(s) - Single Client Access Name Virtual IPs
Node VIPs - One per node
ACFS Registery - For mounting ASM Cluster File System
GNS VIP (optional) - VIP for GNS

Level 4: CRSD oraagent spawns:

ASM Resouce - ASM Instance(s) resource
Diskgroup - Used for managing/monitoring ASM diskgroups.
DB Resource - Used for monitoring and managing the DB and instances
SCAN Listener - Listener for single client access name, listening on SCAN VIP
Listener - Node listener listening on the Node VIP
Services - Used for monitoring and managing services
ONS - Oracle Notification Service
eONS - Enhanced Oracle Notification Service
GSD - For 9i backward compatibility
GNS (optional) - Grid Naming Service - Performs name resolution

This image shows the various levels more clearly: