![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
"I've been a Linux/Unix Admin of sorts since about 1995, and the need for the sorts of sysadmins who don't or can't *program* is declining dramatically.
Back in 2014 I was part of a team of 6 Systems Administrators doing the operations and support on a 450 (approximately) node Content Distribution Network. In 4 *working days* in November we did a bare metal upgrade of *every* server on our CDN, during business hours without taking the CDN offline in *any* area. We did this through a combination of planning, pre-distribution (through scripts) of custom built images, and several other scripts and "home built" tools.
The TL;DR of the following is "Over time the 'Systems Administrator' title will go away, and jobs that have that function will involve programming".
In my not so humble opinion, if you can't program--in at least bash/csh/powershell--you *aren't* a Systems Administrator, you're a computer operator or a hardware guy (Not "don't" program, "can't" program).
With the advent of systems like Docker and Cloud compute engines like EC2 and Google compute--where it's almost easier to treat instances as immutable and redeploy from scratch rather than fix/upgrade existing instances, and with the increasing prevalence of tools like SaltStack, Chef, Ansible and Puppet for both physical and virtual, and the myriad of other tools for doing this stuff, administration of physical machines, by hand is (slowly) draining down. It won't ever go away, but there will be an increasing number of "old" Unix admins who don't make the jump chasing an increasingly small number of jobs.
With the increasing use of stuff like Amazon's Lamba service, ElasticBeanstalk, and the similar "serverless" compute infrastructure as a service offerings from Google, IBM, RackSpace et. al. the will be increasingly fewer new Unix/Linux admins required.
I don't have as much visibility on the Windows side, but I see Microsoft offering Active Directory as a cloud service (Amazon and I assume Google have the same), as well as Exchange/Mail services. So I assume a similar "thing" is happening there.
What this means (I believe) is that what we today call "Sysadmin" (again on the Unix/Linux side) is going to split into a few different areas. There won't always be clean lines between all of them.
- Hardware guys. Mostly doing low level rack and stack, running cable and the like. This won't ever go away, but as machines increase in power and disks increase in capacity there will be MORE virtualization, which means fewer machines installed per "server"/instance and these will be mostly lower paid physical labor positions. Since everything "above" logging the hardware in the CMDB will happen
- SRE/Operations types. These already exist, and it's the current "exit" path for most unix admins. However at the low level it's eyes on glass and responding to errors/outages. As you go up from tier one (moving towards SRE) there is more troubleshooting (mostly of systems), and at higher tiers you are writing code to respond to problems and working with "pure" developers to build more resilient systems.
- These are also the guys that will be responsibe
- Build/Deploy engineers. These will be writing code to build and deploy applications including building and deploying virtual machines/cloud instances.
- Some sort of fusion between #2 and #3 will be the guys who actually handle the stacks that bring physical machines into "the cloud" or otherwise lay the operating systems down on them. They will be a mix of junior folks who push buttons on web interfaces and more senior types who writing code to build and deploy the operating systems.
There will also be internal IT/Helpdesk types to provide help with desktop systems, but I think that desktops are going to go the same way servers are--that central (more or less) systems will build and install them, control them, upgrade/patch them and audit them. This is mostly already the case.
This won't happen overnight--but the transition from Systems Administration as a hand crafted artisanal practice to a practice of complete automation has been going on since 1993 (CF Engine's early days) and picked up serious steam in the early 2000s. I don't expect it to be finished in 5 years, but in 20 years I expect that the requirement for "Systems Administrators" will be about what it was in 1990."
К слову, не так давно (где-то пару месяцев назад) у меня от конторы был однодневный курс Introduction to DevOps. И мы как раз там рассматривали изменения сквозь призму изменения подхода к operations. По сути, курс был сжатым пересказом нескольких книг, таких как "Four principles of Low-Risk Software Releases" by Jez Humble, "Phoenix project" by Gene Kim, "Continuous delivery and integration", "DevOps Handbook" и т.д.
Но главная мысль там была в том, что в целом требования к админам стали намного выше, чем скажем лет десять назад. "Классические" админы вымирают как класс.И девопсы это уже другой класс задач, подходов и оценок. Вплоть до того, что девопсы должны уметь подходить к менеджерскими метриками и оценками к своей работе. Собственно, на этом же курсе мы разбирали реальные workflow реальных компаний и смотрели, какие ошибки были допущены. В том числе и предлагали, что нужно изменить и сверялись с тем, что было реально сделано в компаниях.
Этот курс, а также многие другие вещи сподвигли меня к пониманию того, что мои знания скудны и ограничены и надо учиться. С другой стороны и поле для деятельности открывается огромное.
