https://github.com/brendandburns/kubernetes/blob/hyperkube/d...
And I think I can get it down to a one-liner.
(also, can you update your blog to point to the hyperkube:v0.14.1 image instead of :dev, :dev is a random binary from my client, where as v0.14.1 is an official release... Thanks!)
My biggest worry and issue with running kubernetes in production is the overall workflow of standing up an environment, which seems to be:
1. download some images/dockerfiles 2. [magic] 3. profit
I cannot find a document anywhere that tells me what components are used, how they interact, what settings are required, etc. I'd love to be able to give kubernetes a try and see how it would work for our service, but am having a very hard time getting the right level of detail. It appears to be either "go the [magic] route", or "read all the code" with little in between.
If I had some kind of pointer, I'd be happy to write something up about how to get it running for a prod setup...
This whole area is still in the early stages; any documentation you see on specifics are likely to be soon out of date.
If you try the hyperkube command you'll see very many command line options, and I can only see that growing.
To an extent though, there is a certain amount of magic when downloading images that do stuff for you. For example, the scheduler is a pod that just starts up on the master. What exactly it does, I've no idea yet.
To quote the kub github page:
"Kubernetes is in pre-production beta!
While the concepts and architecture in Kubernetes represent years of experience designing and building large scale cluster manager at Google, the Kubernetes project is still under heavy development. Expect bugs, design and API changes as we bring it to a stable, production product over the coming year."
Maybe this is getting lost in translation, but I find this line to be somewhat condescending. I am asking for some hint as to where I can find documentation that goes beyond "here, run the docker/vagrant/VM image, and have fun" - I want to know what is what, which pieces talk to which pieces, and how. I specifically not asking "Is this ready for production" which is a decision I am happy to make for myself, on the basis of the research I hope to do.
Just to be clear: I am an (admittedly increasingly unfashionable) infrastructure guy. I am pretty sure that if I follow your examples, I can get something up and running that allows me to feed some DSL into some tool, and have a running set of containers. Not interesting to me. I am happy to believe that this works, and am happy to take people's word for it.
I want to know and learn about required infrastructure, about failure domains, about networking requirements, about load and overhead, and ultimately, about "If I build something like this, what are going to be the issues in making sure it will never go down". I see a lot of high-level architecture, which appears to segue quite suddenly into "now do magic, and here is how you then start a pod". I am asking about this in between stage, as I have been unable to find this.
> This whole area is still in the early stages; any documentation you see on specifics are likely to be soon out of date.
I would expect any documentation to be at least somewhat relevant to the version it is released with. It isn't important for us to be on the latest greatest - it is important for us to know and understand how the stuff we use behaves and is to be operated, especially in failure modes. If my only recourse during a failure is "well, lets try to restart a container running something critical and lets see if the problem goes away" then indeed it isn't ready for anything other than being a toy.
I do see the likes of Kismatic and now tectonic making moves to run this as a production system, so somewhere, somehow, it would be possible to stand up a system that has not had all the key decisions made for you, and would allow you to build something that is ready for a particular environment.
Kismatic have actually released some packages that appear helpful in pulling the pieces apart and I will be looking at those to get a better understanding of what does what.
> To quote the kub github page: "Kubernetes is in pre-production beta!
Yeah, I saw that, thanks... From the perspective of the Kubernetes team "pre-production" likely means that they have not yet evaluated many probable edge-cases for many different use-types. This is important to them, not so much for me. What is important for me is that my workload works. It is a lot easier and faster for me to test this (and feed back the results, thus helping the project towards production status) on the basis of an infrastructure I know and understand than it is for me to try and put together something something [magic] and trying to figure out stuff isn't working. Case in point: I followed one of the VMware install guides. at the "now run this image, and xyz will happen" nothing happened. No reasonable possibility of troubleshooting, as the image in question was/is a black box, and no documentation I could find, outside of the "invoke these magic incantations".
To be very honest, I am not too bothered - there are plenty of viable alternatives that do the same / similar things. We are evaluating a large list of possible environments, and kube got crossed off the list pretty quickly, which is a pity as it looks interesting.
Anyway, FWIW I've had similar troubles with documentation with this and OpenShift Origin.