4 ms·

Local Django on Kubernetes with Minikube

41 points by waprin 10y ago

rckrd 10y ago
Not the author of this post, but a maintainer of minikube. Glad to answer any questions.
- jfkw 10y ago
  Is there any plan to support MiniKube with LXD as VM driver (not container engine)?
  - rckrd 10y ago
    Not really. We run in a buildroot [0] based VM on your laptop so that we can run k8s components and a docker daemon cross platform (windows, osx, linux). On linux, it might be an interesting idea, but it looks like theres still an upstream issue on kubernetes to support lxc/lxd [1]. Right now, minikube supports rkt and docker. And the systemd integration of rkt is pretty nifty. [0] https://buildroot.org/ https://buildroot.org/ [1] https://github.com/kubernetes/kubernetes/issues/6862 https://github.com/kubernetes/kubernetes/issues/6862 [2] https://github.com/kubernetes/minikube https://github.com/kubernetes/minikube
- atmosx 10y ago
  I started playing with kubernetes today and used minikube! Thanks for building it!
- foxylion 10y ago
  Is there anything planned to support multiple nodes. So that one can test outage scenarios, load balancing across nodes, etc? If not what other local deployment method do you suggest that works good? (All other methods seem to be deprecated)
  - rckrd 10y ago
    We aren't planning to support multiple nodes anytime soon. The resource overhead would probably be too much for most users just looking to try out kubernetes. You can take a look at the k8s docs about running locally. There's a script that brings up a cluster with a configurable amount of nodes on linux. [0] https://github.com/kubernetes/kubernetes/blob/v1.5.0/docs/devel/running-locally.md https://github.com/kubernetes/kubernetes/blob/v1.5.0/docs/de... [1] https://github.com/kubernetes/kubernetes/blob/v1.5.0/hack/local-up-cluster.sh https://github.com/kubernetes/kubernetes/blob/v1.5.0/hack/lo...
- cwisecarver 10y ago
  Can you outline the differences between minikube and kube-solo?
collyw 10y ago
Django doesn't really seem like a framework that is suited to cluster computing, its almost always backed by a relational database and usually IO to the DB is the bottleneck, assuming you now how to use it. Can someone explain the advantages of running it like this to me?
- subway 10y ago
  Er, the same as the advantages of running any web application on k8s. Django web applications themselves are well suited for stateless hosts. You don't need any persistence where the Django app runs, you push your DB persistence off to another host (or container backed by a persistent volume).
  - collyw 10y ago
    I was under the impression that the advantages were being able to spin up multiple instances quickly, whereas I usually see Django running on a single server. What are the other advantages of running on k8?
- waprin 10y ago
  As the other poster said, you run Django on Kubernetes for the same reason you run anything on it: * Containers are an easy way to package an application so that it's easy to run anywhere, even if it has weird dependencies * Eventually, you'll want to run containers on more than one machine for performance/reliablity * Kubernetes makes it easy to schedule containers on a lot of machines, have them talk to each other over a virtual network easily, and provides a toolkit to solve a bunch of other random things you'll probably want (image updates, secret management, namespaces, authentication, jobs, etc) So the biggest reason to run Django on Kubernetes is you're running other stuff on it, or just because it's one of the many ways to run a containerized app. Sure, you could write a Compose file and just run Docker on a single VM, but you're probably going to quickly want to add other apps or run on more than one machine, at which point you are likely going to look at Kubernetes, Swarm, maybe Mesos/Marathon, or roll-your-own container orchestrator. Running Django on Kubernetes makes a lot of sense. Running something like Postgres on Kubernetes is the part that is admittedly more questionable. It's mostly just done for fun/as an exercise. Long term, databases will probably run in containers since everything else is. Short-term, for a serious project, I would probably just use a managed database service.

41 points by waprin 10y ago

rckrd 10y ago

Not the author of this post, but a maintainer of minikube. Glad to answer any questions.

jfkw 10y ago
Is there any plan to support MiniKube with LXD as VM driver (not container engine)?
- rckrd 10y ago
  Not really. We run in a buildroot [0] based VM on your laptop so that we can run k8s components and a docker daemon cross platform (windows, osx, linux). On linux, it might be an interesting idea, but it looks like theres still an upstream issue on kubernetes to support lxc/lxd [1]. Right now, minikube supports rkt and docker. And the systemd integration of rkt is pretty nifty. [0] https://buildroot.org/ https://buildroot.org/ [1] https://github.com/kubernetes/kubernetes/issues/6862 https://github.com/kubernetes/kubernetes/issues/6862 [2] https://github.com/kubernetes/minikube https://github.com/kubernetes/minikube
atmosx 10y ago
I started playing with kubernetes today and used minikube! Thanks for building it!
foxylion 10y ago
Is there anything planned to support multiple nodes. So that one can test outage scenarios, load balancing across nodes, etc? If not what other local deployment method do you suggest that works good? (All other methods seem to be deprecated)
- rckrd 10y ago
  We aren't planning to support multiple nodes anytime soon. The resource overhead would probably be too much for most users just looking to try out kubernetes. You can take a look at the k8s docs about running locally. There's a script that brings up a cluster with a configurable amount of nodes on linux. [0] https://github.com/kubernetes/kubernetes/blob/v1.5.0/docs/devel/running-locally.md https://github.com/kubernetes/kubernetes/blob/v1.5.0/docs/de... [1] https://github.com/kubernetes/kubernetes/blob/v1.5.0/hack/local-up-cluster.sh https://github.com/kubernetes/kubernetes/blob/v1.5.0/hack/lo...
cwisecarver 10y ago
Can you outline the differences between minikube and kube-solo?

collyw 10y ago

Django doesn't really seem like a framework that is suited to cluster computing, its almost always backed by a relational database and usually IO to the DB is the bottleneck, assuming you now how to use it. Can someone explain the advantages of running it like this to me?

subway 10y ago
Er, the same as the advantages of running any web application on k8s. Django web applications themselves are well suited for stateless hosts. You don't need any persistence where the Django app runs, you push your DB persistence off to another host (or container backed by a persistent volume).
- collyw 10y ago
  I was under the impression that the advantages were being able to spin up multiple instances quickly, whereas I usually see Django running on a single server. What are the other advantages of running on k8?
waprin 10y ago
As the other poster said, you run Django on Kubernetes for the same reason you run anything on it: * Containers are an easy way to package an application so that it's easy to run anywhere, even if it has weird dependencies * Eventually, you'll want to run containers on more than one machine for performance/reliablity * Kubernetes makes it easy to schedule containers on a lot of machines, have them talk to each other over a virtual network easily, and provides a toolkit to solve a bunch of other random things you'll probably want (image updates, secret management, namespaces, authentication, jobs, etc) So the biggest reason to run Django on Kubernetes is you're running other stuff on it, or just because it's one of the many ways to run a containerized app. Sure, you could write a Compose file and just run Docker on a single VM, but you're probably going to quickly want to add other apps or run on more than one machine, at which point you are likely going to look at Kubernetes, Swarm, maybe Mesos/Marathon, or roll-your-own container orchestrator. Running Django on Kubernetes makes a lot of sense. Running something like Postgres on Kubernetes is the part that is admittedly more questionable. It's mostly just done for fun/as an exercise. Long term, databases will probably run in containers since everything else is. Short-term, for a serious project, I would probably just use a managed database service.