Oracle Clusterware Components.

Oracle clusterware is a software which facilitates multiple nodes to communicate
with each other and make them behave as a single Server.

Oracle clusterware is run by Cluster Ready Services(CRS) which in turn is dependent
on two components.

a) Oracle Cluster Registry(OCR)
b) Voting Disk

Also, CRS service itself has four Components.

i)   Cluster Ready Services Daemon (CRSd)
ii)  Cluster Synchronization Services Daemon (OCSSd)
iii) Event Volume Manager Daemon (EVMd)
iv)  Oracle Process Monitor Daemon (OPROCd)


a) Oracle cluster Registry(OCR)

OCR is responsible for recording and maintaining the cluster and
node membership information.

b) Voting Disk

All the nodes continuously send their heartbeat information to the voting
disk.

i) Cluster Ready Services Daemon(CRSd) 

This is the main clusterware daemon which uses the OCR to manage resources
in the cluster.  

CRSd is responsible for starting, stopping and monitoring nodeapps (GSD,VIP,ONS,listener).

CRSd runs as 'root' on Unix and 'LocalSystem' on windows and restarts in case
of failure.

Failure of the CRSd process results in restart of the CRSd process without node reboot.

ii) Cluster Synchronization Services Daemon(OCSSd)

OCSSd is responsible for providing cluster synchronization services between nodes.
It runs as the 'ORACLE' user.

Failure of the OCSS daemon causes node reboot to avoid split-brain situation.

OCSSd is also available in a single instance configuration
if Automatic Storage Management (ASM) is used.

It Provides

Node Membership,
Lock Management and
Group Management services.
 
 
iii) Event Volume Manager Daemon (EVMd)

Is responsible for callouts.

Callouts are specific actions that can be configured to take place
on specific events. It is somewhat similar to triggers.

Callouts can be configured by scripts in the crs_home/racg/userco directory.

iv)  Oracle Process Monitor Daemon (OPROCd)

Oprocd provides I/O fencing, it uses the hang check timer or watchdog timer for
the cluster integrity.

This process runs as the 'root' user.

Failure of the OPROCd process causes node reboot.