A lot of startups reach a point where they look at their perfectly working deployment setup and say:
This isn’t serious enough. We need Kubernete...
Some comments have been hidden by the post's author - find out more
For further actions, you may consider blocking this person and/or reporting abuse
yeah this hits for me - i chased all the flashy stuff early on and just ended up with headaches tbh. makes me wonder, you reckon sticking with simple always pays off or there’s a moment where jumping into complex pays back big time?
My number one concern with CaaS offerings is that these don't offer any network level isolation. So if an application is having 5-6 microservices which should be able to communicate with each other over a private network only, how can one create multiple isolated environments (dev, prod etc) for such application on a CaaS offering?
When it comes to kubernetes, it offers isolation at multiple levels. Cluster is the first level of isolation while second level of isolation can be achieved using namespaces. It has ingress controller which is the only service exposed to the internet while other services may remain private.
Next is how to achieve path based routing to expose all these micro services under the same hostname?
There is a huge gap between the simplicity of CaaS offerings and high complexity of k8s which, IMO, docker could have filled with docker swarm by implementing some of the features of k8s.
K8s is having lot of automation and monitoring built into it. So it not only reconciles the desired state but monitors the containers too to recover from the failures to a great extent which in turn reduces the efforts of devops teams significantly.
Yeah but most dev teams nor companies understand how to manage and run it. Sure all it offers is nice but expensive to run. Start with containers and go actually with the needs of your business instead of making it technically cool (complex)
Running k8s may be expensive but if you consider the money saved by reducing the efforts for (dev)ops teams then you are saving money effectively.
I think the usecase is already complex enough to justify k8s, but then again, do you really need 6 microservices when youre just starting out, or would a monolith work just as well?
Sadly Docker Inc basically completely dropped Docker Swarm once they realised that k8s won the enterprise space. They aren't really interested in improving Docker Swarm anymore and consider it a failure :(
Monolith works if only one person is developing all the services but with k8s I can have a development environment shared among multiple developers where they can deploy their respective services independently without impacting other developers' work.
On the other hand, I agree that people should start with a CaaS offering if it meets their short term goals and move to k8s only when they need some feature which is not possible to achieve with a CaaS offering.
You can still have separate tasks/modules with a monolith :)
Nice post, exactly the problem most companies face. Too complex infrastructure, leaking abstractions, everything is crazy complex for no reason. And people think your app is cooler if it runs on Kubernetes without understanding what it does. Elasticity is nice, when you need it. Usually you know when you need it
Thanks :) I once had a boss who's argument was always "but google does it that way!" (we were 3 devs lol)
You should always use the right tool for the right job and this guy has clearly never used Kubernetes.
i dont use computers, not a fan
Welcam
Use the right tools for scaling. If your app doesn’t need k8s, lighter options are fine, but migrating to k8s later can be painful. It’s often better to start with a comprehensive tool like 8s. You don’t have to fully utilize it from day one, and unlike development frameworks, it won’t hurt performance just by being there.
What do you consider development frameworks?
Interesting to read
Yeah, been there chasing the 'real tools' badge way too early myself. Keeping it boring actually saved me so much time
Overview: Many startups consider adopting Kubernetes due to its perception as a professional tool, but it may not be necessary for their current needs.
Kubernetes Background
Startups vs. Google
Common Misconceptions About Kubernetes
Alternatives to Kubernetes
rsyncandsystemd) may sufficeWhen to Consider Kubernetes
Hidden Costs of Kubernetes
Final Takeaway
made with love by axrisi
sorry but what the actual fuck is this bullshit?
your article in condensed form?
why you call your own article bs?
Do you genuinely think you're adding value with that "summary"? This is pure low effort LLM spam. Please don't do that.
that's your opinion :)
you are free to share it, but not to tell what other people should or should not do
Your comment is clearly against the ToS of dev.to, can they tell you what not to do or is that also not ok?
I have nothing to say, except that world is toxic because of people like you :)
Feel bad for your curstomers if they have to deal with that
Cheers