Hacker News Reader

Comment by icameron | Hacker News Reader

That’s a tough policy to only update prod biweekly! It would be super frustrating if you had a bug crawl out and not be allowed to patch it for 2 weeks. This post really expresses the frustration of working in a bureaucratic environment where developers don’t have full access to change production.

That being said CI/CD is a luxury for coders at lean startups, but there’s still a lot of jobs where you have to work with some DevOps Team to deploy your code to prod. Organizations past a certain size have more hoops to jump through, for reasons.

Of course as a dev it’s ideal to have full access!

View on HN

You know that making CI/CD doesn’t mean you have to pay boatloads of money to a vendor.

Putting up bash script that pulls repo and deploys it is already CI/CD.

Setting up basic Jenkins installation for a technical person should not be taking longer than 2 hours. For person who already is familiar with Jenkins that would be 30mins.

Once you have paying customers I would say there should be max and minimum 2 devs that can fiddle with prod. Others should pass changes via senior people.

A truly lean team (say, <=5 people and limited project scope) should be able to live off their code forge's free CI/CD minutes, or whatever is included in the basic tier they're running. Just run the suite on a schedule against trunk instead of on every PR.

If not, then that's a good signal they should invest more into their CI/CD setup, and like you said it's not necessarily a huge investment, but it can be a barrier depending on skills.

What skill? You can get Jenkins running in afternoon.

If you can't set up CI/CD you're not qualified to program anything

That's a bit harsh, depending on how a person developed or where they worked they may not have had exposure to other facets beyond basic development. Beyond that, it might as well be magic. They'll have to figure out how to provision a VM, ssh into it & lock all the proverbial doors first. Without going into managing it with IaC tools like Terraform, Ansible, Packer, etc.

> That's a bit harsh, depending on how a person developed or where they worked they may not have had exposure to other facets beyond basic development. Beyond that, it might as well be magic.

...so? You sit your ass down and learn. It might take a bit longer if you never touched shell but it's far easier than anything actual programming deals with, especially currently with set of ready or near ready recipes for every environment.

Yes yes. You’re right. I am saying at some places devs don’t own production- there’s an IT/Ops/non-dev person in the loop. Especially common if you’re a consultant in non-tech industries

> That being said CI/CD is a luxury for coders at lean startups, but there’s still a lot of jobs where you have to work with some DevOps Team to deploy your code to prod. Organizations past a certain size have more hoops to jump through, for reasons.

It takes next to no time to setup some basic one and not all that much time to setup decent one, and returns on investments are huge. There is no startup small enough where that isn't a good return.

> That being said CI/CD is a luxury for coders at lean startups,

Really? even before github actions, circleCI did that sort of thing.

Gitlab's runners are nice an easy to configure. Plus all CI/CD is fancy bash with some git triggers.

FT.com used to deploy either directly through heroku to prod, or later via circleCI, and that was in 2015.

Yeah, I can't imagine being a small team building a SaaS and not having 'deploy-on-merge' set up within the first few weeks.

I completely agree. I really worded this poorly. I was trying to say it’s great to have CI/CD to production. There are places I’ve worked who don’t have it due to their bureaucracy/regulation/security. Not because we don’t know how it set it up

Reply on news.ycombinator.com