My dad was never at university, but he was a unix admin for ages. his naming conventions for clusters?
Star Wars characters.
Red Dwarf Characters.
Star trek characters.
Asimov’s robots.
and apparently, his annoying bosses. (For the troublesome clusters.)
I’ve heard it’s a “pets vs cattle” thing. When you have a small fleet of distinct servers, you name them. When you have a thousand interchangeable boxes, you give them systematic IDs.
Or you scale up to a franchise with a large enough cast. I wonder if anyone uses One Piece character names for servers?
It kind of also depends on how you interact with them- some clusters are interacted with by admin as a single entity; those got names even if they technically represented lots of rackspace; or the hardware that’s running specific groupings of services.
Like a databases. (Darth Vader was reserved for databases that logged and tracked errors… aka other systems that were, uh, rebellions.)
You give systematic id’s to completely interchangable things. You give unique names to unique things.
If you name a formal thing (like a physical computer) by its function you have failed at naming. And are probably a manager who doesn’t see that one day you’ll need many things of almost the same function and to tell them apart. Or that one thing will have many functions.
To anyone reading this and not getting it. When your pet gets sick you take care of it (named special servers/other machines). When a cow in the feed lot gets sick you…replace it.
My dad was never at university, but he was a unix admin for ages. his naming conventions for clusters?
Star Wars characters.
Red Dwarf Characters.
Star trek characters.
Asimov’s robots.
and apparently, his annoying bosses. (For the troublesome clusters.)
I’ve heard it’s a “pets vs cattle” thing. When you have a small fleet of distinct servers, you name them. When you have a thousand interchangeable boxes, you give them systematic IDs.
Or you scale up to a franchise with a large enough cast. I wonder if anyone uses One Piece character names for servers?
It kind of also depends on how you interact with them- some clusters are interacted with by admin as a single entity; those got names even if they technically represented lots of rackspace; or the hardware that’s running specific groupings of services.
Like a databases. (Darth Vader was reserved for databases that logged and tracked errors… aka other systems that were, uh, rebellions.)
You give systematic id’s to completely interchangable things. You give unique names to unique things.
If you name a formal thing (like a physical computer) by its function you have failed at naming. And are probably a manager who doesn’t see that one day you’ll need many things of almost the same function and to tell them apart. Or that one thing will have many functions.
To anyone reading this and not getting it. When your pet gets sick you take care of it (named special servers/other machines). When a cow in the feed lot gets sick you…replace it.