How much do you know about server maintenance? Here, I will show you some things I know. First, I hope to give you a basic reference, and second, I hope it will arouse your awareness of server maintenance.
1. Power Control
Although a server is not as fierce as an Induction Cooker, It is weaker than an ordinary PC, and it is also a heavyweight power user.
The most basic point of server hardware application is to achieve operational stability and continuity, while maintaining the operating stability of the hardware system, power stability is the foundation. In this way, when we deploy the power system in the IDC room, in addition to the sufficient supply of the Municipal Power of the server room, we also need to be able to cope with sudden power outages.
Assuming that a data center has 100 servers with an average rated power of 500 watts, a large UPS Power Distribution Cabinet equipped with 96 UPS dedicated high-specification batteries can basically ensure that after the mains supply is stopped, the machine runs normally for 8 hours (ideally, it is about 6.5 hours in practice ).
2. Temperature Control
When most brands of servers on the market run, if there is no other control equipment, the average CPU temperature is above 60 ° C, and the internal temperature of the box is also above 40 ° C, when the concurrency processing is busy, the above two scales may be improved by around 10-20 °C, which is quite different from the theoretical description in the manual. If the server continues to run for an hour at a time of busy concurrency, who knows what will happen next? Therefore, when we build a Server Runtime Environment, we must implement temperature control.
So how can we establish a temperature control environment? Use the air-conditioning system!
If it is an IDC or ISP data center with a large number of servers, you must set one or two centers to ensure that the main and large volumes of central air conditioners are required for daily temperature control, in addition, prepare one or two spare parts of the same specification. If one is an air conditioner with less than 10 parts, at least three vertical or outdoor wall air conditioners with a total horsepower should be prepared (do not use indoor wall air conditioners to avoid serious consequences due to air leakage), and the room temperature should be controlled within 15-23 ° C.
In addition, if it is a large data center, it is best to have a temperature sensor for monitoring.
3. Humidity Control
The humidity control of the server's surrounding environment is also very important.
Assuming that the server runs normally in a dry environment, it is easy to generate static electricity when it is exposed and rubbed around, especially around metal devices. The impact of static electricity on the server is clear to everyone. In case of carelessness, it is easy to cause current breakdown of important components such as capacitors or CPUs. The consequence is not only the system crash, this poses a great threat to the security of the operator.
China's geographical conditions are South Chao and North stem. In the data center in the North, we try to place a humidifier in the data center. In the South, especially in the data center on the first floor, we lay moisture-proof materials under the floor of the large data center, we 'd better place some water-absorbing infrastructure such as lime sandbags to prevent the equipment room from being too wet. The humidity in the data center in the North and South China should be controlled between 45-55%.
In addition, in the rainy days, it is best not to open windows for small data centers to avoid rainwater entering the house, causing unnecessary electric shock in the confidential environment of the data center's power facilities.
4. Fire Risk Control
Many people may think that fire control is irrelevant, because many facilities in the IDC are made of insulation materials. But in fact, there have been many machine room fire accidents. The plug-in board must select more formal, safe and reliable brands. It is best not to place the plug-in board for testing next to the water dispenser. Be careful when working on electroplating and soldering, it is best to go downstairs to solve the problem.
Of course, after solving human factors, we need to deal with unexpected environmental factors. It is best to prepare an independent alarm device for buildings without smoke suppression.
5. Lightning Protection
Electronic devices are very sensitive to lightning sensing. If you do not pay attention to it, it may be dangerous.
Many buildings do not pay too much attention to lightning protection facilities. If the data center is in a building without a lightning rod, it is best to coordinate the property to install lightning protection equipment on the top floor of the building, direct the attacked lightning to the Earth.
6. dust-proof
A server is a high-performance machine and a vulnerable body. Due to the exposure of servers in some data centers to the air for a long time, when the dust in the air enters a certain amount, the fans in the machine may be overwhelmed and the strikes may begin; in addition, the dust enters, for most devices in the host, including the motherboard, the CPU life is very high loss.
Therefore, in the IDC room, it is best to purchase professional server cabinets when conditions are met; before the management personnel enter the IDC room, it is best to put a one-time dust cover or personal clean slippers on the foot; in principle, the IDC does not accept visits from outsiders.
7. Dodge
Direct sunlight is very helpful for the increase of server temperature, but it is a pity that the higher the server temperature, the more prone to problems, and the stability of the server system is very unfavorable. In addition, direct sunlight is very aggressive for the display in the machine room-because of direct sunlight, the life of the display is easily halved or even more.
When operating in IDCs and ISP data centers, it is impossible to see a glimmer of sunshine that represents hope, so it is very good to avoid the light.
In some small data centers, the room facilities are crowded due to the need to make full use of the rental cost of space. It is likely that there is a large window not far away; the environment in the IDC room is very boring for friends who yearn for freedom but have to sit in it for a long time. They cannot tell when to open the curtains and open the windows, feel the sun and the fresh air. It is a good thing for people to feel the sunshine. It is a disaster for machines to feel the sunshine.
Therefore, when the sun can direct the window to the server room, it is best to add a rule, that is, prohibit the opening of curtains and windows. However, the rules are always designed to be user-friendly. Considering the experience of the data center staff, it is best to take several free hours every day to let everyone go out in turn.
8. Pressure Control
Each server has a certain degree of pressure. Although it is a fully metallic body, there is always a maximum pressure. Generally, tower servers are stand-alone and stand-alone. Even if they are horizontally stacked, the number of servers piled up will not be too large because the space occupied by a single server is too large, and the pressure on the external environment involved here is not great.
Generally, for a better server chassis, take a 1U rack-mounted chassis as an example. The actual stress that a 1U can withstand is about 5-7 of the same capacity (that is, 1U; some rack pallets with good strength are generally under pressure between 6-8 1U servers.
Therefore, you must make a budget when setting the Cabinet layout specifications. Do not place too many cabinets in a single compartment.
9. Space Control
The Space Control of the server is mainly for convenience of planning and management. There is also a small reason for the above mentioned temperature control to achieve better heat dissipation.
The messy placement of servers or the display of Network cables are everywhere, and there will be a sense of boredom from the direct sense of vision, when there is a problem with the server or part of the server in one day, we need to handle it!
I personally experienced this situation many years ago. I only felt that the chaotic cabinets and underfloor lines were shaking through the optic nerve! It took more than half an hour to pull out a five-meter-long line because it was a test line without the current line plan and contact point.
Compared with many years ago, I was much better at the subsequent line processing, because the server was placed fully to control space and machine spacing, and the re-planning of machine locations was well organized, line Control is also very sequential, so it is easier to solve the problem.
Good space control is also very good for temperature control, just like a network cable. What if a pile of Network cables are placed in disorder behind the cabinet? Blocking the air flow of the server below, and the temperature increases steadily "!
If it is a tower server built on the wall in a small data room, I would like to make full use of the space, strive to maintain temperature control, maintain power supply and daily maintenance of K \ V \ M and other lines, it is recommended that the wall be 12cm away from the body.
- Unlimited server hardware maintenance (Part 1) Disassembly
- No limits on server hardware maintenance (ii) Fire Prevention and moisture prevention
- Server hardware maintenance without limits 3) Dust Removal