Shulou(Shulou.com)06/01 Report--

In the RAC cluster

Cache fusion-- GCS: implements the replication and transfer of data blocks between instances

-- GES: responsible for managing locks

Lock PCM_Lock and Non_PCM_Lock (to coordinate node values)-Lock Manager DML

-manage two kinds of resources

-decide whether to allow users on the node to request access to a block of data.

-two resources CacheFusion resources and Non_CacheFusion

-CacheFusion resources-exponential data block resources, such as index blocks, blocks stored in common table data, blocks stored in restore segments

Record the version information of the data block, current status, node distribution

-Non_CacheFusion-- such as database cache, data files, parameter files

Lock PCM_Lock is used to coordinate concurrency control and access to data blocks between nodes.

Advantages of RAC: improved system reliability (HA) and system load balancing (LB)

High availability:

System reliability = 1-(1-single point failure rate) ^ nodes

The higher the number of nodes, the higher the reliability of the system.

Balance the system load:

When the system is loaded, RAC can automatically balance the load among multiple nodes, reducing the computing pressure on a single instance.

RAC= database system composed of multiple single nodes, running a single instance on each single node

Special problems of RAC

Concurrency control

Using DLM mechanism to manage concurrency control between instances (shared storage)

Amnesia

Due to the need of daily operation, the database is shut down, and the replica configuration files in the cluster cannot be modified synchronously.

Cerebral fissure

The heartbeat mechanism. The problem of control-voting algorithm

Logical isolation

IO isolation

Problem and purpose: to ensure that logically isolated nodes cannot access shared data, that is, physical isolation

Solution: software method-restart the failed node

Logical and physical structure of RAC Cluster

Storage layer

RAC=n*instance+1*database

Datafile, online logfile, controlfile, etc. are stored on shared storage

Each server is connected to the storage through a HBA card and an optical fiber line.

Instance=SGA+background process

Database=datafile+logfile+controlfile+...

Network layer

The network accessed by the Public network card provides query service to the outside world.

Private network composed of Private network cards, used for heartbeat and Cache fusion of RAC

A storage network consisting of storage devices, fiber switches, and HBA cards of each node

The first two IP packets are transmitted, and the last is SCSI data.

Cluster component layer

Storage device sharing, RAC adds a layer of cluster ware between OS Kernel and oracle

It will intercept the request before OS Kernel, then negotiate with the clusterware on other nodes, and finally complete the upper-level request.

CRS (cluster ready service) must be used on the Linux platform

Application layer

The whole application layer is composed of several CRS Resource monitored by CRS.

Resource is divided into two groups: nodeapps and database-related resource

Including GSD, ONS, VIP, Listener automatically created in the CRS process

Including Database, Instance, Service can be created automatically, or through DBCA, Netca, Srvctl

Detailed explanation of RAC Architecture

The role of 1.GRD

This area is used to store the difference of the same database on different nodes, that is, when multiple instances operate a data block concurrently, they put the data block in the GRC memory area of their respective instances.

2.LMON process

Check their current status by coordinating communications

3.LMD process

LMD provides GES services to coordinate access to the same data block between nodes.

4.LCK process

The locking process, which is responsible for the lock management of unified block access in the cluster.

5.LMSn process

LMSn provides GCS to copy blocks between nodes

6.DIAG process

The "log" process, which records the health status of the cluster and records diagnostic information in the event of an instance error

7. Storage of parameter files and log files

Parameter files and log files must be shared between instances of all nodes.

Clusterware component

1.OCR

OCR is a disk file that is stored on a bare device and is used to store configuration information for RAC cluster nodes

Because there is only one disk file for storing the configuration in the entire environment, synchronous changes to the cluster configuration are implemented.

2.Voting Disk

The file is used to store the node state, and the file is universally stored to the bare device, which can be accessed by each node. Once a node fails, the failed node is proposed to cluster through the "voting" algorithm, and a healthy cluster is reconstructed.

3.occsd background process

The cluster synchronization service uses a mechanism to judge whether the nodes in the cluster are "alive" and monitor the health status of the nodes.

The ocssd process completes the css service.

4.crsd process

The crsd process monitors and manages resources by reading the resource status information stored in the OCR file, and provides high availability of the system in the event of resource failure.

5. Other clusterware processes

The EVMD process is responsible for distributing events generated by CRS

The RACGIMON process is responsible for checking the database status

6. Network Settin

Two network cards. Private+Public

During the clusterware installation, you also need to configure VIP, which is bound to the Public network card to implement the failover of the RAC cluster.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.