Technical naming of servers. What is the principle of naming servers?

Question

Network administration

I want to reflect and listen to your experience.

Background.
Often I had to invent technical names for projects and servers.
They used the names of satellites of planets, chemical elements. I also heard variants of Japanese dishes and cigarette lines, the names of cities. Who cares what.

But I don't think those words fit very well. For example, cities can be unconsciously associated with geographic coordinates, and the "Moscow" server is just a name, it is not physically located here, there may be confusion. For example, if it's part of a CDN. And so you can dig into any word. For example, you are against smoking, and your projects are on the winston server.

It's even worse when the name of the server is associated with some kind of project that will be running on it.
Because the project is growing, new servers are appearing, and it will be difficult to name everything by the name of the project.
And it's just a computer. Virtual machines or dockers from other projects can be added there.
So the name should be without associations, preferably short and even better so that the number of characters in the name is the same on all servers.

Idea
The shortest version of naming I got was 2 characters.
For example (a0.example.com, b9.example.com, xz.example.com)
The bottom line is that you need to draw a table like chess or sea battle or in excel.
The table will have columns from A to Z and fields 0 - 9 - A - Z


This table shows which servers are busy and which are free. If you fill them in order, then you can understand which are new, which are old (a0, a1, a2, ...)

If a person works with a hundred servers, then he will not remember them even if he calls them by the names of relatives.
In any case, he will need some kind of wiki.
If he works with a couple, then C1 and B6 will be remembered as well as the names of their parents.

And if you write a simple program that displays this table, in which you can create projects and bind servers there and give users permission to access projects, then this is generally beautiful to me.

Question
What methodology do you use?
Can I reinvent the wheel?
Maybe there is even special software?
Or is there some fundamental mistake in this approach, the pitfalls of which have already filled you with cones?
Share your experience. I think it will be useful for many to read holivar on this topic.

Answer 1

Well, our servers are called:
CR-MOW-DC (rehabilitation center - Moscow - domain controller)
MS-SYK-HV (Supersky Medical Center - Syktyvkar - hypervisor)
Well, the names are fictitious, but the essence is conveyed.
And by the name I understand which organization the server belongs to, in which city it is, and what it does.

Answer 2

I use naming based on the physical location of the server.
{dts name}-{rack number}-{unit number in the rack}
It turns out something like:
miran-8-40
itsoft-1-1
If you want shorter and fixed length, you can take the first letter of the DC (or two letters, if there are matches) and align the numbers with a leading zero:
m-08-40
i-01-01
If there are not so many servers to rent entire racks, then just the server serial number:
m-01
i-15
Well, or two/three leading zeros, if number of servers 1000/10000:
m-001
i-015

The shortest naming option I got was 2 characters.

Why not 1 character?

Answer 3

Depends on your project and on the number of machines.
If you have a couple dozen cars - whatever you want to call it.
If a couple of hundred - some kind of convention will do.
If thousands - a specific convention, most likely taking into account inventory requirements or location.
You say the server can either docker or move to another rack or something else. This is a very niche situation. Usually a server is bought for a task and it serves it. If it is not needed for this task, it is removed, reinstalled, and possibly under a new name that matches the new purpose.
If the server does not have a specific purpose, just geolocation and branding is used.
ibm-ny-0519-01 - IBM, New York DC, 2019/Mar, 01
hp-sf2-0619-01 - HP San-Francisco DC2, 2019/Jun, 01
PS
But I understood your main problem. For some reason, you are tied to the host name, and you forget that you can add a DNS alias to the host, and more than one.
Therefore, the hostname of the machine must carry the inventory data.
And if a particular application is constantly running on the machine, you add a DNS alias like myapp.mycompany.intranet.com
If another "idea" appears, then another DNS alias is added myanotherapp.mycompany.intranet.com
If the application has grown or requires others resources - it leaves for another host along with its alias, and everyone is happy.

Answer 4

Typically, the naming convention (formalized and approved) uses location + organizational units in the organization + physical\virtual + server roles + two-character numbering as encoding.
Since it is not clear what kind of organization you have, it is difficult to give specific advice.
Here is an example for a multinational company:
2 characters of the zone / part of the world number + 2 characters of the iso country code + 2 characters of the city code + V \ P (virtual physical) + 1 character of the role + two-digit serial number.
Let's get Z1USNYVD01 - virtual domain controller 01, located in the Americas Zone, USA country, New York City.
If there is a division by Tier, 1,2,3 more tiers are added.
Here is an example for an organization within a country - city, office, tier, type, role, serial number.