Kubermesh: self-hosted/healing/provisioning, partial-mesh network K8s cluster

by guifortaine • 9 years ago

95 24 9 years ago

oarsinsync• 9 years ago

As someone who doesn't know enough about K8s, this looks like an amazingly easy way to get stuck in.

Can anyone who has experience with real deployments advise on what pain-points may be encountered by growing something like this further than N nodes (and talk about what N might be?)

raesene9• 9 years ago

If you want to get started with Kubernetes I'd recommend either, Kelsey Hightower's Kubernetes the hard way (https://github.com/kelseyhightower/kubernetes-the-hard-way) which is a good way to start getting an understanding of the various components and how they fit together or, if you want to try out running workloads on a cluster, Kubeadm (https://kubernetes.io/docs/setup/independent/create-cluster-...)

If you're interested in books on the topic, I like Kubernetes in Action (https://www.manning.com/books/kubernetes-in-action) or Kubernetes: Up and running (http://shop.oreilly.com/product/0636920043874.do)

One of the challenges with Kubernetes is that it's pretty fast moving, so it's a good idea to work with resources that are up to date. I know a last commit from May doesn't seem very old but that's going to a least be missing 1.7 and could be missing 1.6 or earlier as well.

meddlepal• 9 years ago

Just to parrot... the governance and development model for Kubernetes is amazing. There are releases every three months and it is like clockwork. That said I usually wait for the first point release to ship before upgrading. 1.x.0 releases always feel like an RC and there's usually some critical bug fixed between 1.x.0 and 1.x.1.

Project plug. I am building a free service and tool for instantaneous cloud-running ephemeral Kubernetes clusters called Kubernaut. If you just want to play around with Kubernetes and start learning it is a great tool. Also useful for CI use via Travis: https://github.com/datawire/kubernaut

dguaraglia2• 9 years ago

I've found a few not-so-obvious pain points on my very limited k8s experience.

1) k8s makes a lot of sense for stateless applications (such as your website) but not so much for stateful applications that require a client to connect to the same container every time (there are ways to do it, but they are a pain in the ass.)

2) Tooling is getting better with time, but it's still pretty green. Packages for your usual orchestration tools like Puppet and Ansible are volunteer work so they get easily out of sync, or require more work than you'd expect to get going. Using their suggested YAML format leads to another problem: there's no easy way to keep secrets outside of the configuration files, unless you build your own process around it

3) Some pieces, like a replicated DB, might be easier to be kept outside of the k8s cluster. You can technically run them there, but they weren't designed for running in that kind of environment and sometimes it shows

4) The CI/CD story using Jenkins pipelines is far from solved. There are several packages that provide some solutions, but the documentation tends to be horrible and that leads to days of debugging through trial and error

5) Leaky abstractions everywhere. As an example, the Jenkins plugins to build using your k8s cluster suggests using the "credentials" system, but you need to add them manually after you boot the Jenkins service. Then your slaves stop receiving the credentials and you have to reboot Jenkins (I had to reboot it on average every 3 or 4 builds.)

Don't let that discourage you from using k8s as a "run" cluster, if your app is a shared-nothing, stateless app. It's so much easier to setup (specially on Azure and GCP) and it obviates the need for setting up Puppet + Sensu + load balancing just to make sure your service keeps running when a node dies.

alexnewman2• 9 years ago

1) That's true, but why not support shared state on shared storage. If that's not an option.

2) What exactly do you not like tooling wise.

3) Totally agree. I am actually working on replacing etcd with postgresql though.

vkat• 9 years ago

For #1 have you tried using local storage available in 1.7, see example https://github.com/kubernetes-incubator/external-storage/tre...

kuschku• 9 years ago

Using their suggested YAML format leads to another problem: there's no easy way to keep secrets outside of the configuration files, unless you build your own process around it

Isn’t that what the Secrets resources are for? Just use those, and mount them into your containers in the filesystem or environment variables.

markbnj• 9 years ago

Regarding CI/CD we use gitlab pipelines to build and push containers and run 'helm upgrade --install' to update charts in the cluster to use the newly pushed images. Works very well and not at all difficult to set up.

web007• 9 years ago

This sounds like buzzword bingo.

I can't tell without reading the source, is this any different from minikube? I see a mention of NUCs and netboot in the README, so I'm guessing it's slightly more, but not documented.

mbryant2• 9 years ago

There's a better explanation of what's going on over here: http://ocadotechnology.com/blog/creating-a-distributed-data-...

There's some libvirt stuff, but that's mostly for local testing without having a bunch of physical hardware.

The key things I'm trying to do are: - Automatically provision new machines, just by plugging them in. - Don't have a traditional network (i.e. switches)

corndoge• 9 years ago

From the article, you're using Quagga -- are you aware of the active fork FRRouting[0]? It was forked about 8 months ago and is over 3,000 commits ahead since most of the core developers have stopped developing Quagga.

[0] https://github.com/FRRouting/frr

jzelinskie• 9 years ago

NUCs are pretty standard for how many people are testing bare metal installations on the cheap. It's what we use in the labs at CoreOS.

This looks like an opinionated installation for an actual bare metal cluster complete with anycast mirrors of popular registries. Minikube is only really good for running in a VM locally.

dghubble• 9 years ago

Just to clarify - there are a few NUCs, but its mostly SuperMicros and other vendors where testing and qa are done.

dominotw2• 9 years ago

kubernetes has so much momentum and developer interest. I rarely see anything about DC/OS which is what they choose at my work last month, apparently its DC/OS is more 'battle tested' in enterprise.

I don't think i even want to setup dc/os on my laptop and play with it, seems almost impossible to run mesos on osx . All these articles are making me resent DC/OS :D.

justicezyx• 9 years ago

DC/OS was developed with objectives that are not overlapping with K8s'.

Namely, they target machine management. While K8s is about container centric job management. Seems subtle, but if you think along these 2 lines, design decisions in 2 systems fall through naturally.

dominotw2• 9 years ago

DC/OS was developed with objectives that are not overlapping with K8s'.

from their home page ' deploys containers, distributed services, and legacy applications into those machines; '

Containers is listed first and its definitely an overlap with kubernetes . I understand that mesos was originally developed with non container centric focus but dc/os now definitely container focused. First example on their site in graphic is 'Containerized Workloads '.

BlackjackCF• 9 years ago

Having worked in DC/OS prior to Kubernetes, I'm going to say that it's much easier to jump into DC/OS development (working with the API etc). However, Kubernetes has been a lot easier for me to get running a lot faster, mostly due to the much more active community.

It also feels like the recent feature sets that Mesosphere has been delivering to Marathon (their choice of scheduler to sit on top of Mesos) is playing catchup to a lot of the features that Kubernetes already has.

Kubermesh: self-hosted/healing/provisioning, partial-mesh network K8s cluster

Related Stories