- Oracle Clusterware is the software that enables a database to be opened by more than one instance.
- In 9i was named Oracle Cluster Manager oracm; renamed in 10.1 to Oracle Cluster Ready Services and finally renamed to Oracle Clusterware(OC) in 10.2
- In 10.2 and above OC installation enhanced by adding the CLUVFY (Cluster Verification Utility) which can be used to check each OC installation stage requirements are met.
- OC must be installed on a separate home other than oracle database home.
- OC consists of three main components:
1- CSS (Cluster Synchronization Services): Used to manage cluster node membership.
2- CRS (Cluster Ready Services): Manage high availability operations.
3- EVM (Event Manager): Manage event notifications and callouts.
On UNIX those components implemented as daemons and entries added to the /etc/inittab to spawns those daemons in case of node reboot.
Before 10.2 the only way to start the CRS was by reboot the node. Starting form 10.2 CRSCTL used to start and stop the CRS.
- OC uses Two additional files that must be accessible by all nodes:
1- OCR disk: Used to store details of the cluster configurations including names and statuses of databases, instances, services and node applications such as listeners and VIP's.
OCR is used by OC and managed by client tools such as DBCA, SRVCTL and OEM.
2- Voting(quorum) disk: Usedto determine the nodes that currently available in the cluster.
Used by the OCSSD to detect nodes that join or departs the cluster. For that its also called css voting disk.
- Both OCR and Voting must be stored on cluster file system or raw devices but not ASM.
- Both disks can be mirrored to eliminate the potential of SPOF.
- One mirror for the OCR and two for the voting.
- Initial size for the OCR is 100 MB and 20 MB for the voting.
Node Applications:
- Those applications are automatically configured on each node during the OC installation.
- They will be automatically started if they enabled when the node is booted. They also can be started and stopped by SRVCTL utility.
1- Listener:
Run on each node to listent to incoming connection requests from the clients.
Can be managed by LSNRCTL and SRVCTL.
In RAC environment its configured on the OCR to run as a node application.
2- ONS (Oracle Notification Service):
Used by OC to propagate messages both within RAC cluster and clients and application-tier systems.
Uses publish-and-subscribe method to genearte and deliver event messages.
ONS provide the foundations for FAN which in turn provides the basis for FCF.
3- FAN ( Fast Application Notification):
Is a mechanism by by which RAC notifies other processes about changes in configuration and service level.
Also used to notify applications about service status changes such as start/stop of instances and services.
FAN publish also load-balancing advisory events. Application can take advantage of those events to direct work requests to the instances that is providing the best level of service.
4- FCF ( Fast Connection Failover):
Its used to prevent new connections to be directed to the failed nodes or instances.
It works with integrated connection pools in the application servers and clients.
Oracle clients that provide FCF are JDBC, OCI and ODP.NET.
5- VIP (Virtual IP):
Alternate IP to the public IP addresses.
have the same subnet as the public IP addresses.
Configured using the VIPCA utility. If a node failed then its VIP address fails over to another another surviving node.
Used to redirect the client connection request to a failed node to another surviving node.
This eliminates the TCP timeout issue which it takes a longer time.
6- GSD (Global Services Daemon):
Allows clients (srvctl,dbca,..) to execute administrative commands such as instance startup and shutdown.
In 9i it was implemented as a java program, managed by GSDCTL and run as a daemon.
In 10.1 and above its implemeted as a node application and can be only managed by SRVCTL.
- In 9i was named Oracle Cluster Manager oracm; renamed in 10.1 to Oracle Cluster Ready Services and finally renamed to Oracle Clusterware(OC) in 10.2
- In 10.2 and above OC installation enhanced by adding the CLUVFY (Cluster Verification Utility) which can be used to check each OC installation stage requirements are met.
- OC must be installed on a separate home other than oracle database home.
- OC consists of three main components:
1- CSS (Cluster Synchronization Services): Used to manage cluster node membership.
2- CRS (Cluster Ready Services): Manage high availability operations.
3- EVM (Event Manager): Manage event notifications and callouts.
On UNIX those components implemented as daemons and entries added to the /etc/inittab to spawns those daemons in case of node reboot.
Before 10.2 the only way to start the CRS was by reboot the node. Starting form 10.2 CRSCTL used to start and stop the CRS.
- OC uses Two additional files that must be accessible by all nodes:
1- OCR disk: Used to store details of the cluster configurations including names and statuses of databases, instances, services and node applications such as listeners and VIP's.
OCR is used by OC and managed by client tools such as DBCA, SRVCTL and OEM.
2- Voting(quorum) disk: Usedto determine the nodes that currently available in the cluster.
Used by the OCSSD to detect nodes that join or departs the cluster. For that its also called css voting disk.
- Both OCR and Voting must be stored on cluster file system or raw devices but not ASM.
- Both disks can be mirrored to eliminate the potential of SPOF.
- One mirror for the OCR and two for the voting.
- Initial size for the OCR is 100 MB and 20 MB for the voting.
Node Applications:
- Those applications are automatically configured on each node during the OC installation.
- They will be automatically started if they enabled when the node is booted. They also can be started and stopped by SRVCTL utility.
1- Listener:
Run on each node to listent to incoming connection requests from the clients.
Can be managed by LSNRCTL and SRVCTL.
In RAC environment its configured on the OCR to run as a node application.
2- ONS (Oracle Notification Service):
Used by OC to propagate messages both within RAC cluster and clients and application-tier systems.
Uses publish-and-subscribe method to genearte and deliver event messages.
ONS provide the foundations for FAN which in turn provides the basis for FCF.
3- FAN ( Fast Application Notification):
Is a mechanism by by which RAC notifies other processes about changes in configuration and service level.
Also used to notify applications about service status changes such as start/stop of instances and services.
FAN publish also load-balancing advisory events. Application can take advantage of those events to direct work requests to the instances that is providing the best level of service.
4- FCF ( Fast Connection Failover):
Its used to prevent new connections to be directed to the failed nodes or instances.
It works with integrated connection pools in the application servers and clients.
Oracle clients that provide FCF are JDBC, OCI and ODP.NET.
5- VIP (Virtual IP):
Alternate IP to the public IP addresses.
have the same subnet as the public IP addresses.
Configured using the VIPCA utility. If a node failed then its VIP address fails over to another another surviving node.
Used to redirect the client connection request to a failed node to another surviving node.
This eliminates the TCP timeout issue which it takes a longer time.
6- GSD (Global Services Daemon):
Allows clients (srvctl,dbca,..) to execute administrative commands such as instance startup and shutdown.
In 9i it was implemented as a java program, managed by GSDCTL and run as a daemon.
In 10.1 and above its implemeted as a node application and can be only managed by SRVCTL.
No comments:
Post a Comment