Lost: checklist to prepare the data center for the summer

Summer that does not kill us makes us stronger .
Old Tsodov Proverb

Summertime and the living is easy ... Sorry, distracted. Summer in a loaded data center is, of course, not about easy living even once, and for those who want to live without losses until September, it is high time to begin preparations.
It makes sense to prepare, even if the last summer passed without failures, and it is important to understand that flawlessly working air conditioners are a necessary but not sufficient condition. Well, yes, first things first.


1. Recall everything
Of course, you have already done this exercise somewhere in October, but take some time and try once again to recall all the exciting cases of last summer: what problems arose, all of them were eliminated, do you have an understanding of how to act in similar situations today and whether the autumn conclusions need clarification. As they say, why step on the old rake, if you can make room for new ones? :)

2. Reality check
Having dealt with past problems, go to the current state of affairs. Do you have enough cold? To answer this question, we uncork the statistics on the power consumption of server and network iron (at different times of the day) and compare the obtained figures with the design capacity for cooling. If the current load is greater than the project load, you can begin to be afraid. Strictly speaking, there are two options here: to reduce the load or to increase the cooling system, up to equipment modernization (installation of more powerful air conditioners / chillers). You can also leave everything as it is and play Russian roulette, but this is, so to say, an amateur option.
If at the level of general figures everything is within, look at the power consumption of individual racks in order to avoid local overheating. Racks that consume twice the average, also need to either unload, or provide additional cooling.
')
3. A Song of Ice and Flame
Attention, cap on the wires: hot and cold corridors make sense if and only if the air between them does not mix. I repeat, does not mix. How do you hear? Does not mix. At all.
Why am I telling you the obvious? Because I constantly come across the fact that this seemingly alphabet is simply ignored.
Now we go to the machine room and check:

• All active equipment, to the last server, must take the air from the cold corridor and blow it into the hot one. With particular care, I would recommend checking network equipment.
• 1 unit ≤ 1 "piece of iron". Even if the equipment is compact, there is still a lot of space in the unit, and you are a very, very careful owner.
• Nothing should prevent hot air from escaping from the equipment into the (hot) corridor: no power outlets and automatic transfer switches, not to mention tools, which are stored in a warehouse and not in a rack.
• All free units must be closed with plugs. And one more time: ALL.
• The topic with the plugs may seem like some third-rate trifle, but all the equipment around such a mini-portal in the hot corridor will systematically overheat. Do you need it?
• Racks must stand close to each other, without "gaps". For the same reason that free units should be closed. And, of course, if you have “diluted” columns in individual rows of racks, make sure that the entire space between the columns and the racks closest to them is also closed.

4. Clean - a guarantee of coolness.
We have dealt with the machine room - let's go and look at the external units of air conditioners / chillers:

• First of all, we check whether the surrounding landscape has changed: new walls, buildings, etc. - all this can significantly affect (in a bad sense of the word) air exchange near the external units / chillers, so it is important to track such changes in time and take the necessary measures.
• At the same time, you should make sure that you have 24-hour safe access to your external equipment (this is a must), and the site itself is well lit, equipped with electrical sockets and running water (if you don’t want to pull wires and hoses in the country house style through window-door corridors) ).
• Go to the actual equipment. If in the heat exchangers for the winter accumulated dirt (as it usually happens), we take qualified specialists on air conditioners - and yes, my. Why are qualified specialists, and not guest workers? Because unprepared guests of the capital can, for example, bend the lamellae.
• For the period of poplar fluff, we plan daily monitoring of the state of external units and are preparing to additionally wash the lamellae as needed.

5. I twist-twist, I want to check
If you have a reserve for air conditioners (and you should have one), it's time to make sure that this reserve is in working condition:

• If the backup air conditioners are usually turned off - turn them on instead of the main ones and check whether they cope with the current load; equipment rotation is generally useful to carry out on an ongoing basis, and not only in the framework of "military exercises".
• If all air conditioners, including reserve ones, work constantly for you, turn off each “extra” one by one and see how the rest feel.

6. MOT & contractor
It is highly desirable to conduct maintenance of refrigeration equipment before the first hot days (in this sense, this April will help everyone). It is also useful to check in advance the validity period of the current service contract so as not to find yourself in the middle of June without a contractor.
And call your contractor: request an unscheduled departure, make sure that a strong team does not run away during the winter and is generally ready to quickly handle requests.

7. Hope for service, but don't make it yourself.
Even if your contractor is still well done, check out the minimum set of parts and tools that you should have on hand in case of urgent repairs:

• Consumables "for every day": freon, oil, automatic, fuses, that's all.
• Parts whose delivery may take several weeks:

- External and internal fans
- Compressors for chillers \ conditioners
- Controller
- Pressure and temperature sensors

8. Leafing through the old notebook
Check whether you have the actual layout of the external units and readable markings on the external units themselves - in other words, make sure that you have a clear idea of ​​which external unit belongs to which air conditioner, and any engineer can figure this out in a minute.
Along the way it is worth checking out:

• marked introductory machines on air conditioners;
• whether there are actual single-line diagrams in the switchboards to which these air conditioners are connected;
• are automata marked in these switchboards (ideally in “N conditioner” format)

9. Do not relax
In the first hot days, closely monitor the monitoring readings: this will allow you to quickly track all the “bottlenecks” that were not identified for any reason at the preparation stage.

Look like that's it.
Refreshing coolness on a hot afternoon!
PS: do not forget to turn off free cooling :)