[mikece]

This article doesn't make any sense to me. DevOps from my perspective is CI/CD defined with a YAML file and using Docker biuld containers or custom GitLab Runners (or the Github equivalent), not something where programming in Python or Golang is needed.

What am I missing here?

[tester457]

Go is good for infrastructure as code and automation since it compiles to native binaries.

Unfortunately the article neglected to mention that, which imo is the most useful aspect of go for devops, since no one wants to deal with runtime dependencies for hundreds of machines.

[jen20]

You're missing what DevOps is - close personal collaboration between development and operational teams to further delivery goals.

[silisili]

The term DevOps has been bastardized so badly that nobody really knows anymore. At my current company, 'DevOps' are what we used to call sysadmins.

When I first heard the term, it was described to me as 'developers do their own operations.' Next it was 'developers work with operations.' Then it was 'operations, but working to support developers.' Now, it mostly means something between sysadmin and terraform wrangler, I think.

[mikece]

When done right, the dev team IS the ops team.

[hedora]

Why not hire ops people to do ops?

All the competent devs I know either hate ops work, are bad at it, or happily transferred to an ops team, because being split between dev and ops meant they could do neither well.

[kazen44]

this is me experience aswell.

Most of the times devs doing ops works ends in a nightmarish setup which is hard to maintain.

There is a reason specialization exists, and having seperate ops folk can save you a tremendous amount of money on resources or/and hardware.

[verdverm]

Who thinks about and manages the shared systems?

[mikece]

I'm assuming dev teams that are responsible for deploying, observing, and maintaining microservices (which obviates the discussion about shared services). For cloud the physical servers/infrastructure is managed by AWS/GCP/Azure; for on-prem k8s the infrastructure team (what older devs used to call "sysadmins") provide the hardware onto which to deploy our convoluted microservices.

[hedora]

None of what you are saying works in practice.

It pushes all the complexity to the interfaces between teams (what used to be a unit test in a monolith becomes a continuous manual operations task), so that no one is empowered to or responsible for making the application work properly.

Then, it explicitly eliminates the teams and mechanisms that used to manage the complexity of budgeting hardware, forecasting demand, etc, which kills margins and also somehow simultaneously forces developers into unexpected load shedding exercises caused by hardware shortage, leading to roadmap slippage and team churn.

On top of that, it ensures that production issues will be harder to debug because the infrastructure team has no visibility into the application, and the development / ops teams have no visibility into underlying hardware issues.

Thankfully, many smaller/growing companies are explicitly rejecting most of this nonsense (since they wouldn’t be able to succeed otherwise).

[berkle4455]

DevOps is what sysadmins do.