A desktop app really needs a screenshot in the readme.
The argument is that there will be an enterprise version to allow that, with "enterprise licensing".
> We collect data that you submit via kontena account registration. We collect data about events that are happening in Lens app: application started, application stopped, logged in, logged out, cluster settings page open, cluster view open, cluster feature X installed, cluster feature X uninstalled, cluster feature X upgraded, terminal opened and kubernetes resource applied. There is no metadata related to these collected. We are not collecting any kind of telemetry data from Lens app: for example we don't know how many clusters you might have, how many nodes you have, what is actually running in your cluster etc.
vs.
> It's the only way for the company behind the product to see how many users are actively using it.
That's a lot of data for the simple task of seeing how many users are actively using it.
It actually works well, it automatically read the contexts from my local kubectl configuration and it's got a dark theme. I'll be properly test-driving this for the following days.
It does feel snappy, and looks good.
One cute GUI utility I've found useful is https://kube-forwarder.pixelpoint.io/.
I've also found k9s nice to use: https://github.com/derailed/k9s
They had an image in the repo, but it wasn't referenced in the readme.
Are we going to have a kubernetes manager manager soon?
If Kubernetes management is a problem, fix Kubernetes. Ultimately, Docker should be fixed so that kubernetes is not needed. After that, operating systems should allow you to simply copy a binary over and run it, statically compiled. This type of distribution should be the norm and not the exception.
We keep piling things on top of other things, and no one seems to care. There are no adults overseeing us, and we soiled our diapers a while ago.
K8s isn't made to fix Docker, but rather the problems caused by having to run quite many micro-services. The system it was inspired by predates Docker by many years (Google's borg).
OP probably knows. my impression was that they picked docker to represent the group...
Not sure why you are seemingly upset or confused by it.
totally agree.
someone asked earlier why docker swarm is not used in production or not even talked or written about? what are the issues and is anyone trying to fix them?
i honestly don't know either.
but i feel that, on top of us piling things on top of things, we are now also running a kind of popularity show for tools and to some extend languages. whereas the most popular tool is claimed the best and the one everybody should use.
you'd get crucified if you did otherwise...
Nobody is crucifying you if you want to use Docker Swarm. You're just being weird.
People are using Kubernetes because (a) it's easy to customise and tailor, (b) there is a bigger ecosystem of components e.g. ingress, (c) it's not controlled by one vendor like Docker is, (d) Google was quick to offer a cloud version of it.
I would have rather we (as an industry) focused on the developing better languages and frameworks to write our software with. Look at Go for example...
With Go I can create a single binary, produce a DEB or RPM and deploy it using Ansible. If I need to update the software I make a change and push it through the CI/CD stack, resulting in another binary, then another DEB/RPM, then an updated set of servers. I then restart a service... so hard /s
If I need to run micro services, then, you know, just run them? They're all statically linked, singular binaries on disk. If I need to run 500 of them, then I just run them. Each gets whatever port it can find and registers its self with (something like) Consul and just like that, I'm DNS query or an API call away from knowing what services I have, where they are and what port to talk to them on.
I believe we should be focusing on making those processes better, not just adding more complex layers on top of simple solutions.
I know K8s has its benefits, but I believe the same benefits can be realised easily enough without the complexities that K8s brings. Obviously I can't outline them here in a comment on HN, but I am working on providing a POC in the near future.
EDIT: and to those down voting me: that's also lazy. Cognitive laziness. Prove me wrong. Also, I don't care about Internet Points, I care about the truth and finding what's true. I wish you the best of luck though.
What's so hard about sticking the binary in a container? That way you don't need to worry about installing it or generating debs/rpms, everything it needs is in the image.
> If I need to run micro services, then, you know, just run them? They're all statically linked, singular binaries on disk. If I need to run 500 of them, then I just run them. Each gets whatever port it can find and registers its self with (something like) Consul and just like that, I'm DNS query or an API call away from knowing what services I have, where they are and what port to talk to them on.
So finding a free port and registering itself with Consul would need to be built into every service you run. Just run them in containers and let k8s handle the service discovery.
I see what you are saying, but k8s really makes the pattern you are describing easier. It also standardizes it so every company is not managing home-grown ansible and consul mash-ups.
Kubernetes runs more than tiny, statically linked Go apps. It also doesn't just run apps. It can schedule a wide variety of workloads in a cluster, managed their persistent storage, apply disruption policies, autoscale pods based on load or custom metrics, it can set up load balancing, it can run batch jobs, it can run time-scheduled jobs, it can control network traffic, it handles DNS lookups... the list goes on.
Kubernetes isn't particularly complex, but what complexity it does have can be explained. Any attempt to provide the same functionality will ultimately reinvent Kubernetes.
What, by suggesting we run a process on its own instead of wrapping it something that isn't required?
