Note: Since writing this post, I joined Google. We have released a feature called IP Aliases that addresses the problem described in this article, and much more.
I recently ran into a problem while using Google Container Engine (GKE), the managed Kubernetes by Google. This lead me to an interesting solution that can be used for a large range of issues you could encounter in Kubernetes.
The Problem I was unsuccessfully trying to have my pods communicate with an application available through a VPN: everything was working as expected from a VM but the pods in the GKE cluster had no network connectivity with the services on the other side of the VPN.
This is the second and last part of my AWS re:Invent recap. Go check out the first part if you haven’t done so already.
In this second post, I will outline the products announced by Werner Wogels during his keynote. You will also find a small opinionated analysis of the impact of each product, based on the current market and ecosystem.
I tagged the really important ones with a [Game Changer] in the title.
Two weeks ago, thanks to my company, Oxalide, I had the chance to attend AWS re:Invent, in Las Vegas. This is the first part of a recap of all the announcements (yes, there are so many things to talk about that it doesn’t fit in a single post). You will also find a small opinionated analysis of the impact of each product, based on the current market and ecosystem.
Update March 2017: EFS now supports a single mount point for a volume, so the setup is now much easier because you don’t have to differentiate between AZs. Take that into account while following this blog post.
Introduction In the last post we saw how to create a production-ready Kubernetes (K8s) cluster on AWS with Kops. Now, let’s see how to use it in conjunction with AWS managed services to host a highly available application: Gitlab.
Introduction Kubernetes is the leading container orchestration solution. It promises to standardize the way you run applications, without worrying if you are running on bare-metal, on a public cloud provider or on a private cloud.
AWS being the leading public cloud solution, it is important to be able to run Kubernetes easily on this provider. In this post, I will show you how to create a production-ready Kubernetes cluster on AWS from scratch.
Docker For Mac has really changed how I work: I now use it for all my linux-related developments. The integration is OS X is really well done and it’s really perfect for a development environment.
The only problem is that Docker For Mac uses a file called Docker.qcow2 that takes more and more disk space as time passes (mine got to 20GB). Deleting images or containers does not decrease the size of this file.
On september, the 13th, at 7p.m (CEST), Gitlab presented its “Grand Master Plan“: the direction that the company and its product will take in the coming months.
Reminders Gitlab is a company that was created to back the product of the same name. Gitlab was conceived as a competitor of Github: a centralized platform for managing Git repositories with a web GUI. Like Github, Gitlab lets you fork a project (create your own copy), and create Pull Requests (called Merge Request - a better name I think), which is a way for you to submit modifications for code review.
Two weeks ago, I gave a presentation at the 42nd Docker Paris Meetup about Rancher. Rancher is a Docker orchestrator and, as such, a competitor of Docker Swarm and Kubernetes. However, it has some other features that make it very interesting and not incompatible with Swarm or Kubernetes.
Here is the first part of the presentation:
a general presentation of Rancher, a first demonstration: creating and scaling an Elasticsearch cluster in Rancher.
Classic network topology Typically, in a classical infrastructure, network architects design multiple networks according to the security wanted for the elements residing in the network. For instance, for a web infrastructure, you could have a front network, where the webservers live, which is accessible from the Internet on ports 80 and 443 ; and a back network, for the database servers, which is only accessible from the front network. This is a fine design, because it allows you to manage security between the networks with firewalls and routing restrictions.