Overview of Oracle clusterware Composition

Overview of the composition of Oracle clusterware (I) the relationship between clusterware and RAC www.2cto.com in a narrow sense, RAC refers to a "Multi-instance, single database" Environment built with "some cluster components". Here the cluster components can be Oracle clusterware, Sun cluster and other clusterware to virtualize a distributed host into computer RAC virtualizes a distributed instance into an instance clusterware to manage hardware resources downward, providing services to RAC (ii) Composition of Oracle clusterware ① disk Files OCR disk and voting disk must be placed on shared storage (1) OCR disk Oracle due to the "forgetful" problem caused by the synchronization of configuration information of the modified Node the OCR disk is used to save the configuration information of the entire cluster. The OCR disk is located at/etc/oracle/ocr. the loc configuration information is saved as "key-value ".

Each node has an OCR disk copy in the memory, this memory is called OCR cache. The content of the OCR disk can be read by the OCR process on each node. Only the OCR process on the master node can read and write to the OCR process on the master node to update the local disk. and other processes whose OCR cache content requires OCR content, for example, CSSD and EVMD are called client process. These processes do not directly access the OCR Cache, but send a request to the OCR process. If you want to modify the OCR content, the OCR process of the node needs to submit a request to the OCR process on the Master node. The physical read/write is completed by the master OCR process and synchronized to all OCR cache (2) voting disks in the voting phase of cluster reconstruction, the node sends the health status of the members that the node knows to the reconfig manager when network heartbeat is unavailable, when the cluster is split into multiple subcluster, determine the number of nodes in the subcluster of the other party and send killblock to implement evictee eviction notification. This is a signal transmitter crsctl query css votedisk. view the location of the voting disk. ② background the process is in the final stage of installing clusterware, it is required to execute root on each node. sh this script will add the following three lines at the end of the/etc/inittab file: h1: 35: respawn:/etc/init. d/init. evmd run>/dev/null 2> & 1 </dev/null h1: 35: respawn:/etc/init. d/init.css d fatal>/dev/null 2> & 1 </dev/null h1: 35: respawn:/etc/init. d/init. crsd run>/dev/null 2> & 1 </dev/null visible, the most important backend processes of clusterware are CSSD, CRSD, and EVMD (1) OCSSD. If an exception occurs, the system restarts CSSD and uses the heartbeat mechanism, monitors the health status of clusters in real time, and provides basic services such as split-brain protection. CSS service has two heartbeat mechanisms: the disk heartbeat of the network heartbeat voting disk of the private network. Note that when you configure ASM, there is also a CSSD process used to support the communication between the ASM instance and the RDBMS instance. To establish a RAC on the ASM, you need to use $ ORACLE_HOME/bin/localconfig. sh delete deletes this component from/etc/inittab because RAC requires that node only has one CSSD Process (2) CRSD requires HA components, all are registered to the OCR disk in the form of CRS resource, and the CRSD is responsible for monitoring the running status of the CRS resource Based on the content in the OCR, and start, stop, and failover resources (3) the EVMD has two tasks: ● responsible for releasing various events generated by CRS ● A bridge between CSSD and CRSD ③ each node of the network component has at least two NICs to provide connections ● A total of NICs: providing external data services ● private NICs: it is used for communication between nodes. The communication content includes network heartbeat and cache fusion. Each node requires three IP addresses: public IP, private IP, and virtual IP (also known as VIP)