Ms SQL Server 2000 administrator manual series-12. Microsoft SQL Server and Microsoft cluster services

12. Microsoft SQL Server and Microsoft cluster services
Fault Type
Overview
Sample Cluster System
SQL Server Cluster settings
Beyond the MSC
Summary
In recent years, Microsoft SQL Server has moved from desktop systems and work groups to enterprise backup systems. To cope with business operations, these systems become larger and more important. In other words, they must be more stable, easier to manage remotely, and have better fault tolerance. To meet these needs, Microsoft has invested a considerable amount of time and effort to reduce software errors and improve system support. Microsoft has also improved management tools and remote management capabilities, and has developed new technologies such as Microsoft cluster services (MSC. A cluster refers to a group of computers that can act as standby systems when a fault occurs. In this chapter, you will learn how to operate, how to set up, and how to plan to restore the old view from a disaster. You can't fault tolerance your system with the help of MSC alone. You must work with careful planning to make your system have the ability to reply from errors.
________________________________________
Description
MSC is included in Microsoft Windows 2000 Advanced Server, Microsoft Windows 2000 datacenter server, and Microsoft Windows NT 4.0 Enterprise Edition.
________________________________________
Fault Type
　
As a Database Manager, the main task is to maintain the database and enable the database to operate normally within a specified period of time. These are usually stated in the scope of service consent form. The scope of service consent form may also specify the optimal execution duration required by the system, the execution rate, and the amount of time to recover when an error occurs. The use of MSC can increase the maximum execution time and reduce the time required for restoration. Although server hardware, Windows 2000, Windows NT, and SQL Server are generally stable and reliable, some components may still fail. In fact, a complex computer system may have the following types of faults:
• Although the disk technology has been improved, the hard drive is still a mechanical device and will wear out. Hard Disks are one of the most frequently-occurring components.
　
• Hardware component faults may be caused by component wear or damage. The culprit of such problems is the inevitable high temperature in the system. Even computer devices with the best process may fail.
　
• Software Component faults some software flaws are discovered only under specific circumstances. The system may have been running for several months or years before a specific problem is identified. In addition, the addition of applications to a stable environment may also modify key libraries or files, thus causing a fault.
　
• An external fault system may also cause errors due to some external factors, such as sudden interruption of power supply. Whether your system is able to recover from such errors depends on whether you are using a continuously powered system (uninterruptible power supply, UPS) or other surplus power supply.
　
• Human Error clusters often fail to protect systems that fail due to human error. These human errors include accidental deletion of a data table or deletion of a Windows NT file system partition.
　
Errors are inevitable. How to Make the best preparations for these errors is the focus of our discussion in this chapter.
Overview
　
MSC is a built-in service for Microsoft Windows 2000 Advanced Server, Microsoft Windows 2000 datacenter server, and Microsoft Windows NT 4.0 Enterprise Edition. As previously described, a server cluster refers to a single system where an independent server cluster is a group and operates together. The purpose of a cluster is to allow clients to continue to access applications and other resources when an error event or an expected interruption occurs. If a server in the cluster cannot be used for any reason, the resources and applications will be migrated to other nodes in the cluster.
When talking about cluster systems, we generally use the word High Availability instead of fault tolerant ). Traditionally, the term "Fault Tolerance" is more suitable for specific systems that provide extremely high level of backup and recovery. These systems typically use extremely specialized software that enables instant recovery from any single hardware or software error. Fault-tolerant systems are more expensive than Fault-Tolerant Systems. Relatively, a cluster system that provides high availability is not as costly as a fault-tolerant system. The cluster system is generally constructed by the standard server hardware and the cluster-aware software in the operating system. When the availability requirements of the system increase, the cluster can easily join the system as a node. Although a cluster system cannot guarantee continuity, the cluster system indeed improves the availability of most mission-critical applications.
MSC has the following advantages:
• High Availability System resources (such as disks or IP addresses) are automatically transferred from faulty servers to servers that provide services. This process is called failover ). When an application in the cluster fails, MSC automatically starts the application on another server that provides services, or distributes the work of the faulty server to other nodes. The error transfer process is very fast, and the user notices that the service suspension on the server will be extremely short.
　
• Fault-Tolerant responses when the faulty server is repaired and back online, the MSC automatically rebalance the workload in the cluster. This is called "failback ).
　
• Manageability Cluster System Administrator software allows you to manage the entire cluster as a single system. In the cluster system administrator, you can drag the Cluster Object to move the application to different servers in the cluster. You can also use the same method to move data. This drag operation can be used to manually balance the server load, or remove the load from a server for maintenance. The cluster system administrator can also allow you (from any location on the network) to monitor the status of the cluster, each node, and all resources. Figure 12-1 shows the cluster system administrator window.
　
• Scalability when the system needs and loads increase, You can reset the MSC to support these needs. If the overall load exceeds the cluster capacity, you can also add nodes in the cluster.

　
 
Figure 12-1 Windows 2000 Cluster System Administrator
Basic Concepts
　
The reason why the system downtime can be reduced is that the system can provide the error transfer function between multiple systems. To achieve the error transfer function, you must interconnect the servers, and use the shared disk system, as shown in figure 12-2. You can use any high-speed online system, such as an Ethernet network or other network hardware, to connect servers. The server interconnection system acts as the communication channel between servers. This channel is used to transmit the current status information and configuration information of the cluster back and forth. The shared disk system allows the databases and data files of all servers in the cluster to be accessed in an equal manner. Shared disk systems can be SCSI, fiber channel SCSI, or other dedicated hardware. The shared disk can be a separate disk or RAID system (the RAID system has been discussed in Chapter 5th ).
________________________________________
Note:
If the shared disk system is not fault-tolerant and your disk subsystem fails, the MSC will be mistakenly transferred to another server, but the new server will still use the faulty disk subsystem, therefore, your system is still in the fault state. This type of mechanical device is prone to failures, so you must use RAID to protect your disk.
________________________________________
Once your system is set as a member of a cluster server, it will change from a traditional server to a virtual server. Virtual Server looks like a common server, but the entity identification method of the system has been abstracted. Because the computer hardware that makes up a virtual server may be constantly changing, at any time the user cannot determine which server the current application actually runs. Therefore, a virtual server provides services for users' applications, but does not mean a specific set of hardware.

　
 
Figure 12-2 Windows 2000 Cluster
The Virtual Server exists on the network and is assigned a TCP/IP IP address. This IP address will switch between servers in the cluster system. Therefore, no matter which hardware is being executed, the user can see the virtual server. The IP address is actually migrated from one system to another. For the external world, it can maintain a constant virtual server. Therefore, if an error occurs on the server, an application that uses a specific IP Address can continue to access the IP address, even if the IP address currently represents a different server. The Virtual Server prevents the user from discovering the transfer error. Therefore, the user can continue to work without knowing what happened behind the scenes.
Cluster Components
　
Creating a cluster requires several components: cluster management software, server Interconnection System, and shared disk system. These components must be configured and used together with the cluster-Aware Application to create a cluster. In this section, you will learn about these different components and how they work together to build a cluster. In the SQL server cluster Settings section later in this chapter, you will learn how to set the SQL server cluster.
MSC Cluster Management Software
　
Cluster management software is actually a set of software tools that can be used to maintain, set up, and start a cluster. It is composed of sub-components in the lower column that work together to maintain the functionality of the cluster and perform error transfer when necessary:
• Node Manager maintains the relationship between the members of the cluster system and sends heartbeats information to other cluster members (nodes). If the heartbeat of a node stops, the other node will think that it is no longer available and start to take over its work. Node Manager is one of the most critical parts of a cluster because it monitors the status of the cluster and its members and determines what action to take.
　
• Configuration database manager maintains cluster database settings. This database retains tracking information for all components in the cluster, including abstract logical elements (such as virtual servers) and physical elements (such as shared disks ). This database is similar to a Windows NT/2000 logon database (Registry ).
　
• Resource Manager/failover manager starts and stops the MSC. Manager/failover manager receives information from Resource Monitor and Node Manager, such as lost nodes and new nodes.
　
• Event processor initializes the cluster and routes event information between cluster components. Event processor is also responsible for initializing the cluster system by instructing Node Manager to add nodes.
　
• Communications Manager manages communications between nodes in the cluster. All nodes in the cluster must constantly communicate with other nodes to maintain the cluster's operational functions. If each node loses contact with each other, the cluster status information will be lost and the cluster will not work properly.
　
• Global Update manager notifies all nodes of the cluster status, including the addition and deletion of cluster nodes.
　
• Resource Monitor monitors the status of various resources in a cluster and provides statistics. This information is used to determine whether the cluster should take any wrong transfer action.
　
• Time Service ensures that all nodes in the cluster can report the same system time. If there is no time service, the sequence of events may be wrong, resulting in an incorrect decision by the system. For example, if the node that obtains an older version of the archive considers the time to be two o'clock P.M., And the node that obtains the new version of the archive considers the time to be ten o'clock A.M, the cluster mistakenly considers the files in the first system as up-to-date.
　
Server Interconnection System
　
Server interconnect refers to the connection between nodes in a cluster. Because the nodes in the cluster need to constantly communicate with each other through time service, Node Manager, and other components, it is very important to maintain high-speed online.
The server interconnection system must be a reliable channel for communication between systems.
Server interconnection is usually a high-speed Ethernet network that runs TCP/IP or NetBIOS. This setting is enough, but you may want to use a dedicated, high-speed Interconnection System that is faster than the Ethernet. These interconnect devices can be purchased from hardware vendors, some of which also provide communication services such as shared disk services. The Compatibility hardware list for the Microsoft website http://www.microsoft.com/hcl/ provides a complete list of general server interconnect devices.
Shared Disk System
　
Another key component created by the cluster is the shared-disk system. If multiple computer systems can access the same disk system, the system can be moved to other nodes when the primary node fails. A shared disk system must allow multiple computer systems to access the same disk-in other words, each computer must be able to access all disks. In the current version of MSC, only one system can access the disk at a time, but in the future, multiple systems will be allowed to access the disk at the same time.
There are already several types of shared disk systems, and new disk technologies are constantly evolving. The SCSI disk subsystem supports multiple initiators. With this feature, you can have multiple SCSI controllers on the same SCSI bus, so that SCSI is suitable for cluster use. In fact, SCSI is the first disk subsystem used for clustering.
Recently, some new disk technologies have been designed to support clusters, such as fiber channels and some specialized solutions. The optical fiber channel system allows computers to connect to a disk over a long distance. Most Fiber Channel Systems Support multiple controllers on the same fiber trunk. Some raid control cards have been developed or improved to support clusters. If you do not modify or change the settings, most disk control cards cannot support the cluster.
If the cache is in the control card itself, the contents that can be written in the control card cache will be saved in the memory, which will become a major problem for the cluster, as shown in 12-3. In this case, each node has its own cache, which is called before disk sharing because the two caches share the same disk. When the system fails, because each control card has a cache, but the cache is located in the faulty system, the data in the cache will be lost. Therefore, if we use the Internal Controller cache in cluster settings, they should be set to read-only mode. (In some cases, this setting may reduce the efficiency of some systems .)

　
 
Figure 12-3 controller cache before disk sharing
The solution to shared disk system problems involves raid and the cache of the disk system. In this setting, all nodes share the cache, which we call behind the disk sharing, as shown in 12-4. The raid mechanism and the cache of the disk sharing system will be accessed by all the control cards in the system, and the Read and Write cache is safe.

　
 
Figure 12-4 controller cache after disk sharing
Newer SCSI and fibre channel disk systems allow raid control cards to be encapsulated in disks rather than in computer systems. This system provides excellent execution performance and fault tolerance capabilities. In fact, many raid systems provide backup control cards and caches. Many newer raid systems use this structure. The following are some detailed examples.
I/O subsystems have previously mentioned that several different types of I/O subsystems support clusters. The following are the three most important types:
• SCSI jbod is a system that uses jbod (just a bunch of disks) addressing to have multiple controllers on the SCSI bus. In this setting, the disk has an independent address. You must use Windows 2000 or an independent addressing method to set the address to the disk array. This system is generally not recommended.
　
• Each internal raid server has a RAID Controller. The disadvantage of this subsystem is that the controller cache must be disabled to avoid data loss.
　
• The external raid control card is shared by the disk system in the cluster. The cache and raid logic are encapsulated in the disk to connect a simple host bus card (HBA) with an external controller.
　
The next two sections only describe the two raid solutions. SCSI jbod is used only when the cluster is small and the cost is the primary concern.
The internal raid control card is designed to handle the raid process by hardware and the cache belongs to the host system. When internal raid is used, the shared disk system is located after raid, as shown in Figure 21-5.

　
 
Figure 12-5 internal raid control card
Because the cache is on the control card and is not shared, the data in the cache cannot be used when the system fails. This is a big problem in the relational database management system (RDBMS. When SQL server writes data to a disk, the data may still be in the cache (because the operation is to write data to the disk controller cache first ), however, this data is recorded in the transaction record file as written. If the system fails at this time, SQL server will not reply to the data blocks when trying to reply, because SQL Server considers the data has been written to the disk. In this type of error event, the database is damaged.
Therefore, the vendor must ensure that the cache function for the cache raid control card for the cluster has been disabled (or at least the write cache function has been disabled ). If the cache function is disabled, SQL server will not receive a signal that the write operation has been completed before the data is actually written to the disk.
________________________________________
Description
SQL Server writes data to the disk without using the buffer or cache mode. No matter how many File System caches are available, SQL server does not use them. Like most RDBMS products, SQL Server completely ignores the file system cache.
________________________________________
In some cases, the cache of the control card can greatly improve the execution efficiency. Especially when you are using raid 10 or RAID 5, these raid layers impose additional burden on write operations. To use the control card to write data to the cache in cluster settings, you must use an external RAID system to share the cache so that the data will not be lost during error transfer.
External raid in an external RAID system, raid hardware is outside the host system, 12-6. Each server has an HbA, which is used to send more I/O Requests Out Of The RAID system as quickly as possible. The RAID system determines the actual data storage location.

　
 
Figure 12-6 external raid Subsystem
The external raid subsystem is also known as a cabinet raid or a chassis raid because raid is performed in the disk drive cabinet. The external raid subsystem has many advantages. It is not only an ideal solution for MSC, but also an important comprehensive solution. Cabinet raid has the following advantages:
• Easy wiring when using internal raid, you need to connect many cables from the RAID Controller-one for each cabinet. If you use external raid, you only need one cable to connect the HBA to the external RAID Controller, and the Cabinet can be connected in the form of a daisy chain, the Controller is connected to the chrysanthemum ring, as shown in 12-7. Even if hundreds of disks need to be connected, external raid can still do this simply.
　
• Allowing the cluster cache to use external raid makes it easier for you to set a cache raid solution. If you use external raid, you can have both cache and fault tolerance functions without worrying about cache consistency between multiple control cards (because there is only one cache and one control card ). In fact, writing data to the cache is not safe if an external raid is used. Although you are still at risk when caching RDBMS data, this risk is greatly reduced when using an external RAID Controller. Make sure that your external RAID system vendor supports the cache image. When the memory chip fails, the cache image can provide the fault tolerance function.
　
• Supports more disks in large or high-performance systems, and sometimes a large number of disks must be planned. Chapter 1 describes the necessity of a large number of disks and describes how to plan the size of the system. Chapter 2 describes general execution problems and then displays them. The external RAID device allows you to connect hundreds of disks to a single HBA. In an internal RAID system, each controller can only connect to a limited number of disks, just like SCSI.

　
 
Figure 12-7 Internal raid wiring vs external raid Wiring
Among the disk subsystems currently available to support clusters, external raid cabinets are most suitable for large clusters. Of course, the cost must also be considered. In addition, some clusters are too small to use external raid. However, in the long-term implementation scheme, the external raid solution can provide the best performance, reliability and management for your cluster.
Cluster Application Type
　
Applications executed on the System of the MSC can be divided into four categories:
• Non-cluster-aware applications (cluster-Unaware applications) do not interact with MNS. Although they can be executed smoothly under normal conditions, they may be taken away when a fault occurs and cannot force the error to be transferred to other nodes for further work.
　
• Cluster-aware applications (cluster-aware applications) are aware of the existence of MnS. They can use MSC to improve their efficiency and adaptability. They respond well to cluster events and can continue to execute only a few actions (or even none) after component failure and error transfer occurs. SQL Server 2000 is a good example of cluster-aware applications.
　
• Cluster Management applications are used to monitor and manage the MSC environment.
　
• Custom Resource types applications provide custom cluster management resources for other applications, services, and devices.
　
MSC Mode
　
You can run SQL Server 2000 cluster support and MSC in different modes. In active/passive mode, a server maintains the standby mode and is ready to take over the work of the primary server immediately when the system fails. In active/active mode, each server executes a different SQL Server database. No matter which server has an error, the other server takes over it. In this case, a server executes two databases. This section describes the advantages and disadvantages of each mode.

　
 
Figure 12-8 Application Type and MSC
Active/passive cluster
　
Active/passive clusters use primary nodes to execute SQL server applications and use the servers used for secondary nodes as backup or backup servers, as shown in 12-9.

　
 
Figure 12-9 active/passive clusters
In this setting, a server is actually not used. This server may not have been called for any work for several months. In fact, in most cases, backup servers are never used. Because the secondary server is barely active, it looks like an idle expensive device. Because the server cannot be used to execute other functions, you may need to purchase other devices for the purpose of service users, which makes the active/passive mode more expensive than you think.
Although the cost of the active/passive mode is high, it also has advantages. In active/passive settings, if the primary node fails, all resources of the secondary node can be used to take over the activities of the primary node. If you are running an application with a key task and the system needs to load the specified traffic or response time, this reliability is very important. In this case, the active/passive mode may be the right option for you.
It is strongly recommended that secondary nodes and primary nodes have the same hardware (that is, the number of Ram resources is the same, and the type and quantity of CPU resources are the same ). If the two nodes have identical hardware, you can determine that the execution performance of the secondary system is almost the same as that of the primary system. Otherwise, performance loss may occur in an error transfer event.
Active/active Cluster
　
In an active/active cluster, each server executes an application and serves as a secondary server of another node, as shown in 12-10.

　
 
Figure 12-10 active/active Clusters
In this setting, each server is both the primary node of some applications and the secondary node of some other applications. This is the most cost-effective setting, because no device is idle and waits for errors from other systems. Both systems can serve users. A separate passive node can also be used as a secondary node of several major nodes.
One disadvantage of active/active settings is that after a fault event occurs, the burden on secondary nodes increases, resulting in a significant reduction in the efficiency of surviving nodes. The surviving node not only needs to execute the original application, but also the application from the main node. In some cases, if you cannot tolerate the loss of performance, you must use active/passive settings.
Sample Cluster System
　
In this section, let's take a look at four sample cluster systems that use MSC. These examples help you decide which type of cluster best meets your needs and environment.
Example 1-High Availability System with Static Load Balancing
　
This system provides high availability for multiple applications on the cluster. However, if only one node is left online, the efficiency will be greatly reduced. This system has the highest hardware resource usage, because each node can be accessed. Figure 12-11 shows how this type of cluster is set. It is an active/active cluster.

　
 
Figure 12-11 high availability cluster with Static Load Balancing
Each node in the cluster notifies the network of its resources as a virtual server. Each node retains the additional processing capability, so you can execute applications on other nodes when an error transfer occurs. In other words, with these additional resources and processing capabilities, the service that provides the client for the faulty node will not be interrupted.
Example 2-Hot Swap system with maximum availability
　
This system provides maximum availability and performance through all system resources. The disadvantage of this setting is that there is a lot of investment in hardware resources-and most of them are not used yet. One node is regarded as the primary node and meets the requirements of all clients, while the other node is idle. Idle nodes are actually used as standby systems for heat engines. They are accessed only when an error occurs. If an error occurs on the primary node, the hot replacement node can take over all operations immediately and continue to provide services to the client. Figure 12-12 shows this setting method.
This setting is not suitable for some applications with the most critical task nature. If your company is selling on the internet, your web Transaction Server may work in this way. Because your business must rely on the success or failure of the system, it is worthwhile to spend more money to purchase an idle system for emergency purposes.

　
 
Figure 12-12 hot replacement systems with maximum availability
Example 3-local server cluster
　
MSC is a very elastic cluster solution, which can be seen from the cluster settings of local servers. In this system, only the application you selected Allows error transfer. As shown in 12-13, you can specify that some applications are still available when an error occurs on the node, but not on other applications.

　
 
Figure 12-13 local server cluster
This setting method is ideal when you plan to maximize the usage of hardware resources, but must provide limited error transfer capacity for some mission-critical applications. In addition, this setting provides error transfer for cluster-aware applications and supports non-cluster-aware applications.
Example 4-a virtual server cannot be transferred by mistake
　
Our last system example is actually not a real cluster, but it still uses the support of MSC for virtual servers. This setting is actually a method for organizing and releasing resources, as shown in 12-14. The virtual server function allows you to use meaningful and descriptive names for resource naming, rather than a list of common server names. In addition, after the server error is fixed, the MSC automatically restarts the application or resources. This feature is particularly useful for applications that do not provide an internal mechanism to restart. This architecture can also be called the best preparation action towards a cluster. Once you define a virtual server on a single node, you can simply add a secondary node without changing the server definition.

　
 
Figure 12-14 failed migration of virtual servers
SQL Server Cluster settings
　
After you install and set up the MSC, the next step is to set the SQL Server to be executed in the cluster. As mentioned earlier, SQL Server 2000 has the cluster-aware capability and can make good use of the cluster function. In this section, we will first look at how to plan the cluster system, and then review the steps required to set SQL server in the cluster.
________________________________________
Description
To obtain all the benefits of MSC, applications must be cluster aware. As we have said before, cluster-aware applications must understand the cluster architecture and be able to transfer errors in error events. Not all applications support clusters and are capable of "escaping from the cage" when a cluster encounters problems 」.
________________________________________
Plan your settings
　
The first step to plan an SQL server cluster is to determine the type of hardware to be used and the job mode to be used by the cluster. The hardware architecture that can form a cluster system is quite a variety, and the operation of the cluster can choose active/passive mode or active/active mode. The selected mode determines the required hardware type and quantity.
Active/passive cluster settings should be composed of two identical systems, each with the ability to master the overall workload. Because the active/passive mode does not use secondary systems during normal operation, and the primary system is not used after an error occurs, the efficiency of the virtual server will remain unchanged. When a major system error is transferred to an identical secondary system, the user will not notice any change in performance.
Active/active cluster settings should be composed of two systems that execute a specific workload. If an error occurs, the surviving system takes over the workload of the wrong system. In this case, a single system requires two jobs, and all users are faced with performance reduction issues. If you plan carefully, the system's performance may be able to maintain an acceptable minimum limit, but in fact this performance is not safe. When planning active/active cluster settings, you should be prepared to cope with the loss of efficiency, whether to reduce some services or warn users about the error transfer event, the system performance may decline.
The next step to set the SQL server cluster function is to check or change the settings of some SQL servers. In the next three sections, we will check these settings.
Set recovery time
　
When adjusting SQL server performance, you may have set the recovery interval (recovery interval) of the configuration option to another value, instead of the default value 0. Changing this setting can increase the checkpoint interval to improve efficiency, but it also increases the recovery time. In the cluster system, the default value 0 should not be changed. It indicates that the recovery interval will be automatically set by SQL Server. (Having a system that can be mistakenly transferred is the primary reason for using MSC, which should be more important than performance considerations .) This setting allows the database to have a checkpoint almost every minute, and the maximum restoration time is about one minute.
________________________________________
Related information
For more information, see books online and index recovery interval options.
________________________________________
________________________________________
Description
The checkpoint operation allows SQL Server to write the modified data in the cache to the disk. If any modified data is not written to the disk due to a system error, SQL Server will recover the transaction that has been recognized and the transaction that has not been recognized at the start of SQL Server, and restore the unapproved materials.
________________________________________
Set SQL server to execute on active/passive clusters
　
To create active/passive cluster settings, you may have to modify a setting value in SQL Server. If your secondary server and primary server are identical, you do not need to modify them. However, if the secondary server has fewer resources than the primary server, you must set the SQL server configuration option min server memory to 0. This setting instructs SQL Server to allocate memory based on available system resources.
________________________________________
Related information
For more information, see books online and Index Server Memory options 」.
________________________________________
Set SQL server to execute on active/active Clusters
　
In active/active cluster settings, you must set the SQL server configuration option min server memory to 0. If this configuration option is set to manual, the SQL Server may fail to allocate memory after the error transfer. As Windows 2000 is a virtual memory system, it can allocate more memory than actually available. In fact, this is also a common problem, because it causes paging actions. For example, if each SQL server system allocates 75% of the system memory, the merged SQL Server Service requires 150% of the available memory after an error transfer. If dynamic allocation is not possible, the system may be stuck.
Install SQL server in a cluster
　
The process of installing SQL server in the cluster is similar to that described in Chapter 7th. Before starting the installation process, you must determine the location where SQL Server is to be installed. You shall install the SQL Server File on a shared disk controlled by the primary server. The installation path of SQL Server and the installation path of the master database should also be set to point to the shared disk. You should also specify the network communication protocol to be executed by the cluster. Next, follow these steps to install the SQL Server Error migration cluster:
1. Put the SQL Server 2000 optical disk into your optical drive. If your system supports automatic playback, the main Setting Dialog Box of SQL Server 2000 appears, ranging from 12 to 15. If no, You can manually execute autorun.exe (which can be found in the root directory of the optical disk ).
　
 
Figure 12-15 SQL Server Settings dialog box
2. if you have not installed a required OS update suite, or you have not installed a required version of Microsoft Internet Explorer, or you just want to check the list of required installations and click the necessary installation for SQL Server 2000, the installation dialog box for SQL Server 2000 is displayed.
Click on the most suitable system to view its needs, and click on the options you need to enter the installation. If you have loaded the most appropriate software, go directly to step 3.
________________________________________
Description
Before you start to Install SQL Server, you must have installed the MSC.
________________________________________
3. Click the SQL Server 2000 component to see the welcome screen of the SQL Server 2000 installation wizard. If you are running other Windows programs, disable them. Click Next to continue the installation process.
4. On the computer name screen, click the virtual server and enter the server name, 12-16. Click Next.
　
 
Figure 12-16 computer name
5. the user information screen is displayed. Enter your name and company name. Click Next.
6. Software authorization consent. Follow the instructions below to accept the consent form and continue the installation process.
7. On the installation screen, enter the product identification code of 25 characters (marked in yellow outside the optical box ). Click Next.
8. The cluster migration error is displayed, ranging from 12 to 17. Enter the IP address of the virtual server, and click Add. Supports the sub-network address of the MSC. Click Next.
9. view the cluster definitions provided by SQL Server 2000 on the cluster management screen. By default, the local computer is set as the priority node. Other available nodes will appear in the additional node block. Click Next.
10. On the remote information screen, enter the logon credential of the remote cluster node. This logon credential must have the system administrator permission for the remote node of the cluster. Click Next.
11. In the execution individual name screen, select a preset execution individual or specify a named execution individual. To specify the name of the execution individual, clear the default check box and enter the name of the name. Click Next.
________________________________________
Description
The execution individual cannot be named as default, MSSQLServer, or other keywords reserved as SQL Server.
________________________________________

　
 
Figure 12-17 "Incorrect cluster migration"
12. Select the type to be installed in the installation type screen. The installer automatically sets the default value to the first available cluster disk resource in the selected group. However, to specify disk resources of different clusters, click the Browse button under the data file and specify a path on different shared disk resources. Click Next.
13. The Verification Mode dialog box is displayed, 12-18. The setting on this screen determines the security level of your SQL Server installation. You can select Windows Authentication mode or combination mode (Windows Account Verification and SQL Server Account Verification ). If you select Windows Authentication mode, the permissions of database users inherit from Windows user security settings. If you select a combination mode (Windows Account Verification and SQL Server account verification), you can individually define and manage the security of database users. If you select a combination mode (Windows account authentication and SQL Server account authentication), you must enter and confirm the SA (SQL Server System Administrator) logon password. You can also leave a blank field in the SA password field, but this will seriously reduce the security of your SQL Server.
　
 
Figure 12-18 verify Mode
14. In the start copy file screen, click Next.
15. The authorization mode screen is displayed. You can select either the SQL Server client Authorization mode or the server authorization mode or the base authorization mode. In each server authorization mode, you must specify an individual client access authorization (CAL) for a specific server. Each authorization allows one access to the server. The maximum number of client computers that can be connected to the server at any time is equal to the number of client access authorizations that you assign to the server. If you select to authorize each server, you must specify the number of concurrent online client access authorizations purchased for this server.
In each base authorization mode, each computer that is about to access the SQL Server server needs a Client Access authorization. Once authorized, a computer can access any machine running SQL Server 2000 on the network without additional charges.
If you are not sure which authorization mode to select, click each server. This authorization form gives you the opportunity to change to the authorization mode of each base in one way.
Click continue to install the SQL Server application and data files. The SQL Server installer installs necessary files to your system and sets required components. The installation may take several minutes or longer, depending on the speed of your system.
16. When the installation is complete, select the Restart computer option and click Finish.
As you can see, it is quite easy to set up an SQL server cluster. Once a cluster is set, no other setup steps are required. The client accesses SQL Server through an IP address. This IP address will be re-allocated to the secondary node after an error transfer. The program issues to be considered will be discussed later in the section "Beyond MSC.
Use a three-tier Application
　
Most applications are directly connected to the database. The application initiates a transaction and the database responds to the transaction. In a system error event, when the transaction is canceled, the application may also fail to run. In most cases, there is no problem with this setting-if the transaction is not completed, you want the application to fail. However, if you use an error transfer cluster, the database will return to the available status in a short time and continue to respond to the transaction after the error. You can carefully design a layer-3 (three-tier) application to ensure that the application can use this fast recovery service.
In a layer-3 application, the intermediary layer can detect that the server has stopped responding, wait for a while, and then resubmit the transaction. The user may feel a delay in completing the transaction, but the delay is always much better than the failure of the transaction. To achieve this goal, the application must be aware of the connection failure between the server and the application, and be aware of the need to reestablish the connection. The application should also notify the end user of the current processing status using a message dialog box or other means.
Using a three-tier Application to build a seamless error transfer environment is no longer out of reach. Applications must have cluster-aware capabilities and know that virtual servers will soon be able to return to work. By combining a three-tier application with the MSC architecture, You can provide application and data robustness.
Beyond the MSC
　
We have introduced the basic concepts of MSC and how SQL Server works in this architecture. We have also seen how SQL server can survive some hardware and software errors and re-execute transactions in a short time. To achieve this degree of fault tolerance, you not only need to use MSC, but also need to take some measures. There are two steps that are quite important: one is regular backup and the other is to prepare a disaster recovery plan. The system backup program and how to prepare a disaster recovery plan will be described in Chapter 32nd and 33. Backup is not performed if you have a cluster server or a RAID storage system. In many cases, if your system is faulty but no backup is made, these technologies cannot help you. These situations include the following errors:
• Although hardware faults are not common, some hardware faults can make the data look different. If your primary system encounters a hardware fault that is sufficient to damage the database, the secondary system will still use the damaged database after the error transfer.
　
• Software faults, no matter how good the software is, may still contain bugs even if they have been developed and tested as well as possible. If a rare software bug damages the database, the database cannot be used after the error is transferred. Although RAID technology can provide a fault-tolerant copy, the data in the copy may still be damaged.
　
• Users with artificial Errors often mistakenly delete their data. Neither cluster technology nor raid can solve this problem.
　
In Chapter 32nd and chapter 33, you will learn more about contingency plans and how to save the system from difficulties. The previous example only shows the fact that clustering and error transfer serve a specific purpose, but in our war with the "catastrophic, these two weapons only provide stable data access and data integrity. If you want your system to retreat from the crisis, you need more efforts.
Summary
　
In this chapter, you have learned about different cluster settings, hardware and software required for cluster creation, and the SQL server cluster setting process. You have also learned that the status of MSC is indeed helpful, but it is not a complete and complete system fault tolerance solution. You must also use a disk subsystem with fault tolerance and perform the backup. By combining the MSC with a good disaster recovery strategy, the system can provide the optimal system execution time and system reliability. In the next chapter, we will begin to introduce the new functions of transact-SQL and SQL Server 2000 available in SQL Server 2000.