Principles, faults and cases of server maintenance, updates and upgrades

Upgrade principle

1. The purpose of server upgrade is to solve the bottleneck of a certain aspect of the system, but the result of the upgrade cannot cause new system bottlenecks.

When upgrading, you should consider which aspect of the server is the bottleneck. However, from an application perspective, you cannot simply upgrade a certain component to cause other performance bottlenecks. To give a simple example, a single-CPU server can be configured with 256MB of memory, but if a dual-CPU server is configured with 256MB of memory, the memory will definitely become the bottleneck of the system.

2. The premise for server upgrade is that the system hardware configuration can provide users with room for upgrade.

Server upgrades are performed on the premise that the server hardware configuration has the upgrade capability. Therefore, in the early stages of network planning, network administrators should have an initial estimate of application demand and growth. When selecting a model, prepare for future expansion. Leave some space.

3. Upgrades must be targeted at specific applications rather than simply adding hardware.

Server applications are often dedicated to dedicated machines, and one server is mainly targeted at one application. In this way, when a performance bottleneck occurs on the server, the upgrade should be for this application rather than for a certain component. It cannot be assumed that simply upgrading a single piece of hardware can meet the performance requirements of applications.

Upgrade faults and their solutions

Upgrades can be divided into online upgrades and offline upgrades.

1. Online upgrade: The system can be upgraded to a certain component without shutting down, and the system will not be affected in any way. On IA architecture servers, hot-swappable hard disk upgrades and RAID level migrations do not require a system restart.

2. Offline upgrade: The system needs to be shut down, and the upgrade needs to be performed without power. This mainly refers to the upgrade of CPU and memory. For example, when upgrading the CPU, whether it is expanding the CPU or replacing the CPU, it is recommended to reinstall the system. When expanding the CPU, especially if a single CPU is expanded to multiple CPUs, the system needs to reinstall the SMP kernel. For UNIX systems, the system kernel needs to be upgraded and recompiled. For Windows systems, the operating system needs to be reinstalled. Another example is memory upgrade. After the memory is increased, system performance parameters and database application parameters must be adjusted.

Upgrade case

A single-CPU database application server (supports dual-CPU systems). As the amount of data increases and the number of clients increases, the system needs to be upgraded. The Unixware operating system is installed. How to proceed?

According to the user's application analysis, many components in the system need to be upgraded, such as: CPU, memory, hard disk, etc. In this case, the upgrade should include the following steps:

1. Back up system data (to prevent system data loss caused by misoperation).

2. Turn off the power, upgrade the hardware offline, and install the CPU, memory, and hard disk. Pay attention to ensure that the hardware specifications are consistent with those of the original motherboard hardware.

3. Restart the operating system and increase the CPU license.

4. Use the features of RAID to integrate the newly added hard drives.