Golang vs. Python: The Ultimate Battle in DevOps
andrewodendaal.com
andrewodendaal.com
I hope we quickly arrive at a social standard where posting AI-written content is viewed as extremely disrespectful to your audience.
EDIT: based on previous submissions it looks like aoms is the author of this post. Aoms, am I misunderstanding something here? Why are you doing this?
What am I missing here?
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).
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.
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.
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.
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.
“Python is strong because of x. On the other hand Golang is strong because of y. Conclusion: it depends. “
It looks like a template article where *devops* could be replaced by any other topic like: cm/s, e-commerce, microservices or whatever.
As someone who has done devops-y stuff with both, I highly prefer Go just because I can plop in one binary + config file and it'll just work.
With Python I need to install Python, the virtual env manager du jour and all the required packages. And I need to keep them up to date. And everything is just one step harder if I don't, for some reason, have root privileges on the environment in question.
The Go version has been running for days in the time the Python one has all it's dependencies in order =)
I'm more sysadmin than devops, but how fast you can script over something for a quick automation is the defining trait for most devops tasks, not performance. Going for performance is closer to dev than to ops. Using APIs of several systems doesn't need so much performance neither, unless you are doing pretty intensive amount of calls (and the backend/network could be more of a limitation than the language you use in your side).
Yes, the performance and one-file-deploy are Go advantages against python performance (not for that much) and dealing with dependences and versions. But that is not the whole story, at least for most of the tasks I had to deal with.
And it is not "or". You may want to decide between python or i.e. perl for scripting language used for automation, but there are things that should be addressed with python, and others with go.
And there are other languages that ought to at least be considered for scale and concurrency: Scala, Rust, Erlang …