Mail server erection - dual hot standby solution

Double hot standby is a backup technology made by the temporary failure of the server. It is used for hot standby to avoid long-term service interruption and ensure long-term and reliable service. This solution will introduce the solution of the hot standby of the mail server in detail. Take the TurboMail mail system as an example. You can refer to this reference.

1. Cluster Technology

Before learning about hot standby, let's first understand what cluster technology is.

Cluster technology refers to a group of independent computers that use a high-speed communication network to form a computer system. Each cluster node (that is, each computer in the cluster) is one that runs its own process. Standalone server. These processes can communicate with each other and form a single system for the network client that works together to provide applications, system resources, and data to the user and manage them in a single system. When a client interacts with a cluster, the cluster acts like a standalone server. The starting point of computer clustering technology is to provide computer systems with higher availability, manageability, and scalability. A cluster consists of multiple servers with shared data storage space, each communicating with each other via an internal LAN. When a node fails, the application it runs will be taken over automatically by the other nodes.

Among them, a high-availability cluster with only two nodes is also called dual-system hot standby, that is, two servers are used to back up each other. When one server fails, the other server can assume the service task, so that the system can automatically provide external services without manual intervention. It can be seen that dual-system hot standby is the simplest one in cluster technology.

2. Hot Standby Applicable Objects

The general mail server is going to work for many years, and in order to work, the mail backup work is absolutely indispensable. Some enterprises use RAID technology and data backup technology to avoid data loss caused by server failure. However, data backup can only solve the recovery after the system has problems; and RAID technology can only solve the problem of the hard disk. We know that both hardware and software problems will cause disruption of mail services, and RAID and data backup technologies can't solve the problem of avoiding service interruption.

To recover the server, and if there is a slight problem or strong technical support, the server will be interrupted for a period of time. For some users who need to be online at any time, losing the mail is equivalent to losing money, and the loss can be large or small. This kind of user is very difficult to endure service interruption. Therefore, it is necessary to use hot standby to avoid long-term service interruption and ensure long-term and reliable service of the system.

Therefore, whether you need to adopt a hot standby solution, you can judge according to the requirements of the users of your own mail system.

3. Implementation scheme

Two-system hot backup has two implementation modes, one is based on the shared storage device, and the other is the way of no shared storage device. It is called pure software.

1)Based on shared storage devices

Two-node hot standby based on storage sharing is the most standard solution for hot standby. In this way, two servers (the mail system running on both servers) are used, and the shared storage device disk array is used (the data of the mail system is stored in the disk array). The two servers can adopt different methods such as mutual backup, master-slave, and parallel. During the work process, the two servers will provide services to each other with a virtual IP address. According to different working methods, the service request is sent to one of the servers. At the same time, the server detects the working status of another server through a heartbeat line (currently using a private network). When one server fails, the other server makes a judgment based on the heartbeat detection situation and switches to take over the service. For the user, this process is fully automated and completed in a short time, so that it does not affect the business. Because of the shared storage device, the two servers use virtually the same amount of data and are managed by dual-machine or cluster software.

Advantages: For sharing, the database is placed on a shared storage device. When a server provides services, it reads and writes directly on the storage device. When the system is switched, another server also reads the data on the storage device. It provides fast switching in an unattended situation, ensuring no data loss.

Disadvantages: Increased investment in expensive storage equipment, which is a priority for powerful companies.

2) Pure software mode

pure software way, through the mirror software, the data can be copied to another server in real time, so the same data exists on both servers One, if one server fails, you can switch to another server in time.

Advantages:

1 Avoiding a single point of failure of the disk array: For dual-system hot backup, it is to prevent the service due to the failure of a single device. Interrupted, but the disk array just formed a new single point. (For example, the reliability factor of the server is 99.9%, the reliability coefficient of the disk array is 99.95%, and the reliability coefficient of the pure soft dual machine is 1-99.9%x99.9%=99.99%, and the disk array-based hot standby The reliability factor of the system will be slightly lower than 99.95%

2 Saving investment: no need to buy expensive disk arrays.

3 Not limited by distance: two servers do not need to be SCSI Cable length limitations (Fibre Channel disk arrays are also not limited by distance, but the investment is much larger). This allows for more flexible deployment of servers, including increased distance through physical locations.

Disadvantages:

1 Reliability is relatively poor, real-time data replication between two servers is a relatively fragile link.

2 Once a server is interrupted, it will be more complicated after recovery. The data is recovered synchronously. Moreover, the system is in an unprotected state during this period.

3 There is no transaction mechanism. Since the replication is performed at the file and disk level, the success of the replication will not affect the database transaction operation. Therefore, there are cases where the data is incompletely changed. This is quite risky.

4. Configuration

Before setting up the mail server hot standby, you should clearly define your company's needs. Need to set up dual-system hot standby, which way to build, what operating platform to use, specific server configuration, if you use a shared-based way to build what disk array should be used, what kind of cluster software to use, etc.

Hardware: Two servers with the same configuration. You can select

disk array according to your actual needs, and use it to build a dual-system hot standby system.

Software: Mail System

There are many mail systems in the market, the advantages and disadvantages are different, you can choose according to your needs, there are free mail system Sendmail, Qmail, etc., international brands Exchange, Lotus, etc., the quality is not said, But the price everyone knows, there are domestic such as corlmail, turbomail, which turbomail has an exclusive feature, with mail review function ( There are very few mail servers on the market that have this feature), and mail archives are integrated into one system (other mail archives on the market are separate systems), you can experience it.

Cluster Software :

Building dual-system hot backup is of course the cluster software. Here we recommend cluster software heartbeat, this software includes heartbeat monitoring part and resource takeover part, heartbeat monitoring can be carried out through network link and serial port, and support redundancy The remaining links, which send messages to each other to tell the other party's current status. If the message sent by the other party is not received within the specified time, then the other party is considered invalid. In this case, the resource takeover module needs to be started to take over. A resource or service on the other host.