News | April 20, 2015
With so many changes happening in the container space, it was a pleasure to be able to host last week’s London Kubernetes Meetup and gain insight into some of the exciting developments in this area.
Myself, and the team here at OpenCredo believe that Kubernetes is a strong contender in the emerging container orchestration/scheduling space, and with talks from Google, Red Hat and Kismatic lined up we were able to get information directly from key contributors to the project.
Thanks to the organisers Peter Idah and Milos Gajdos, who managed to secure a great line up of speakers:
As a brief overview, we were treated to a deep dive into how Kubernetes works, how we could monitor the platform, and how we might scale our applications. Below is a round up of each talk, along with links to further information.
The night started with a surprise appearance by Patrick Reilly, CEO of Kismatic. Kismatic are pitching themselves as ‘the Kubernetes Company that isn’t Google’, and they are contributing some great tools to the project. Patrick and several of his team have worked on the soon to be released Mesosphere Data Center Operating System (DCOS) for the Apache Mesos cluster manager, and are bringing some of their learning from this project over to Kubernetes. Patrick briefly explained what Kismatic were up to, and also showed us a demo of some very cool looking Kubernetes cluster health metrics graphs:
It’s definitely worth keeping an eye on Kismatic, as Patrick was hinting at lots of cool stuff coming down the pipeline. You can register at the Kismatic website to be kept up to date with information, and as a bonus for visiting the site you will also be treated to an amazing time-lapse video of New York, which if you are as big a fan of NYC as I am is a real treat (and is making me look forward to my visit to QCon New York in June all the more!)
Next up on the mic was James Strachan, software engineer extraordinaire at Red Hat. James ran us through a great presentation of the building blocks (or ‘sub-atomic’ particles) of Kubernetes. For anyone who missed this, most of the information can be found in the Kubernetes documentation, albeit in not quite such a digestible form as presented by James. James stated that he has been developing software on top of Kubernetes for the last six months, and has really enjoyed the experience.
He described the development experience as ‘awesome’ (just a few times!), and mentioned that the framework provided by Kubernetes almost became invisible as he developed tools and applications that took advantage of the platform. This is a good thing, and to James it demonstrated that the framework has been developed as simple as possible, without compromising on functionality.
James also showed us several videos of the tools he had been working on. This included Red Hat’s fabric8, an open source integration platform for management of Java containers (JVMs) and Docker containers, and OpenShift v3, the open cloud Platform as a Service (PaaS) that supports Kubernetes natively. Both of these tools appeared very interesting, and the videos demonstrated the ease at which applications and supporting data stores and middleware could be deployed at scale via a simple GUI with minimal fuss.
James suggested that although the current version of Kubernetes is not yet production ready, now is a great time to start experimenting so that when Kubernetes is released as Generally Available (GA) sometime in the summer, you will be able to hit the ground running.
James’ slides can be found here: http://fabric8.io/presentations/kubernetes-london-openshift-fabric8/#/
The final talk of the evening was provided by Jerzy Szczepkowskia, software engineer at Google. Jerzy has been working with Kubernetes for some time, and gave a great tour of the potential ways in which applications and the underlying platform could be scaled.
Scaling an application deployed on to Kubernetes in a horizontal fashion can be as simple as increasing the number of associated pods in the application’s replication controller, and scaling vertically can be achieved by simply allocating more resources to the associated pods.
Scaling the underlying Kubernetes platform is slightly more difficult, but can be achieved (at least when deploying on to Google’s cloud Kubernetes Engine (GKE) platform) using cloud concepts such as instance groups and autoscaling.
Jerzy’s slides can be found here: http://www.slideshare.net/craigbox/autoscaling-kubernetes
Craig Box from Google was also in attendance, and he shared some very interesting links. Craig mentioned that v0.15 of Kubernetes has been released; that a very interesting Kubernetes Cluster Federation proposal has been put together, named Ubernetes; and that Google have finally released details of their ‘Borg’ cluster manager, from which Kubernetes has emerged (and lessons were learnt!). Cheers for the info Craig!
The night was great fun, and there was lots of interesting learnings and cool things to take away (and experiments to run!) All of the speakers were very informative, and I had some great chats during the breaks with fellow meetup members. Thanks also to the Kubernetes meetup organisers, Milos and Peter, for arranging such a great event. Although Kubernetes is very much an emerging technology, some people really do have some excellent use cases and examples already. I’m seriously looking forward to the GA release of Kubernetes!
OpenCredo were proud to host and sponsor this Kubernetes meetup event, and we are working extensively with Kubernetes, Mesos, Docker and associated technology. We have helped several clients to deploy Docker-based applications into production, including building continuous delivery pipelines and configuring essentials such as logging and monitoring.
Please do get in touch if you have any questions or would like a free initial consultation in regards to Kubernetes and associated container-based technologies.