 |
|
|
Cluster SoftwareOnce the nodes are physically interconnected, either with a cable or switch, you should plan to install the necessary cluster software on all the nodes. The clusterware installation method differs from vendor to vendor and from platform to platform. Cluster software is usually referred to as cluster manager or operating system dependent layer (OSD) and is provided by the hardware vendor. The notable exceptions are Linux and Windows. For example, the following table (6.2) shows the cluster software products that are available for setting up a RAC system. Table 6.2 Cluster Software (OSD) As noted in figure 6.2, Oracle provides the clusterware for Windows 2000 and Linux clusters. Please note that throughout the book, when we refer to ‘OSD’ we mean the cluster related operating dependent system layer. The HA framework supplied by the vendors is not needed for a RAC environment. Instead of using the HA software, Oracle uses its own cluster software or an OSD supplied by the vendor to control the cluster lock management Clustering software monitors and manages the cluster components that includes the servers and the server connections. The cluster software specializes in detecting failures through software probes, fault isolation, and fault containment, wherever applicable. Usually the failure detection process involves determining the fault and applying the response. It could result in either switching over to a secondary component or the total shutdown of that component, in which case some kind of reconfiguration takes place. As an example, if two interconnects are configured, the fault or loss of one would not cripple the cluster, but would merely result in shutting it off. Whereas when one of the nodes hangs, the cluster manager evicts the failed node and informs other members so that the application (database instance) can take the necessary reconfiguration action. The cluster manager is also responsible for node monitoring, which includes polling the status of various resources like interconnect hardware / software, shared disks, Oracle instances, and public networks. The cluster manager interacts with the global cache service in RAC to reconfigure the failed instance resources. Cluster software protects against the conditions of split-brain and amnesia. As noted in an earlier section, split-brain condition occurs when communication between nodes is lost and the cluster becomes partitioned into sub-clusters, each believing that it is the only partition. Amnesia occurs when the cluster restarts after a shutdown using cluster configuration data older than at the time of the shutdown. The changes in cluster membership drive the cluster reconfiguration sequence that may, in turn, result in services being migrated from failed or faulty nodes to healthy ones. On Linux clusters, Oracle provides Cluster Management Software (OCMS). OCMS is part of Oracle9i Enterprise Edition for Linux. OCMS manages cluster membership, maintains a global view of clusters, monitors cluster nodes, and performs cluster reconfiguration. OCMS is installed automatically when RAC is selected and consists of the following components:
With the release of 9.2.0.2, NM and CM are merged into a single service called oracm, and also ‘watchdog’ is no longer used. A full description of oracm (Oracle Cluster Manager) is covered in a later part of this section. For more information, see the book Oracle 11g Grid and Real Application Clusters - 30% off if you buy it directly from Rampant TechPress . Written by top Oracle experts, this RAC book has a complete online code depot with ready to use RAC scripts.
|
|
|
