DevOps Gathering 2018 – Day 1

Hey, how you doing there? I’m back here after a while. Actually I was hoping to write here more often but it’s not happening yet at least. Ok, let’s start this new post while I’m sitting here at DevOps Gathering 2018 in Bochum and came here this morning from Bonn. Maybe we can have another post for the recent CNCF white papers

The talks were just started a few minutes ago by a talk from Viktor Farcic ( Linkedin, Twitter and his career Blog) who has written DevOps like this one here and is member of Docker Captains. He talked about self-adaptation and importance of resiliency and basically you gave an overview to whole infrastructure parts like continues integration, continues deployment, logging. He mentioned tools like Elastic Stack, Prometheus, Jenkins, Traefik, Kubernetes, Docker.

Next we are going to have Joao Santos talking about CI/CD at Zalando. He gave an overview about how they are organizing their infrastructure and teams and how they have provided isolated environments for their teams. Basically he said that each team has it’s own Jenkins instance on AWS and has it’s own k8s clusters on AWS. Each team has one k8s cluster for testing purpose and one for production. During the Q&A he mentioned that they have open-sourced a few tools on their Github account like how to set up a k8s cluster on AWS. He also showed us their usage of SonarQube in their Jenkins’s file. Just find some cool stuff in their Github account: Swagger and ZooKeeper. Also seems that they patroni repo is quite famous. This one was quite fast and brief! 😉

Ok, so I’m waiting for the next one. Nicolas Byl ( Twitter, Blog ) from codecentric AG is going to talk about Stateful Applications on Kubernetes. He was focusing on Storage and volume types we can have on k8s, scaling challenges and its architecture. He also mentioned the option of having external storage on k8s which is here. He was providing a solution for scaling of the volumes which are going to be used by databases within pods, actually right now I guess it might face us to detailed and dangerous race-conditions which are very hard to discover. I also think that this solution might have some conflicts with database replications / clustering techniques. I didn’t understand the solution’s use-case for databases but I guess it might be very useful for CDNs. Nice thing he did he mentioned the resources 🙂 : StatefulSet BasicsPersistent Volumes and his demo files to download ( backup link ). During the Q&A, some nice stuff came up like Vitess which is a database clustering system for horizontal scaling of MySQL, Heptio and its Ark project ( Github ) which helps manage disaster recovery for your k8s cluster, Helm and KubeApps.

Coool, we have a very fancy lunch … :p. While waiting for the next part it’s good to note it here that, It’s spreading like a trend, using AWS services for preparing the infrastructure. They have lots of services and I think it might be good to have a list of them here. Just found this white paper from AWS which seems quite cool and detailed to have general overview. While this presentation and this article from 2017 are also a good resource. Although it might not directly be related to current’s talks here but I guess still having an image here can be more useful and worth it:

An intimidating look at Amazon Web Services
An intimidating look at Amazon Web Services

Now Mario Kleinsasser ( Linkedin, Twitter, Xing, Blog, Github, Keybase ) is getting ready for his talk with the title of Docker: Ops unleashed. So gave an overview about what they had done in their team of 3 persons and some advantages of working in small team like the time you can save in conversations. He said Microsoft and Linux are getting close to each other from the container view. Time is essential, don’t waste the time and remove manual works. make automation and use your time to learn more. They were using OpenVZ for linux containers and then they have switched to docker. It is very recommended to read the book The Phoenix Project, I also heard this from another person too, seems to be interesting. Mario started to show a demo and explain their open-sourced project named Bosnd. The have used confd for configurations in their project. During the demo he mentioned a website which we can use for playing with docker here ;).  He then showed us a short clip from Mailüfterl (maybe can be found here) and it’s English translation name was “truth is what works”. He mentioned the storage problem that still exists in our current solutions. Maybe RexRay can solve this problem in future by providing an abstract layer to communicate with storage types and their specific data structure. They are currently solving their storage problem with an old solution named NAS. I just remembered a good provider for this purpose and it’s QNAP which is described by their plans here.

Next, by having a small change in plans, we had Thomas Fricke ( Xing, Github ) from Endocode AG with the topic of Architecture Patterns for Microservices in Kubernetes, who had also a talk in DevOps Conference ( here ). His talk was quite interesting for me. He started by speaking about tiles and patters, mentioned the Tomb of Hafez and showed the picture of the patterns has been used in its architecture. Then he moved to the role of tiles in DevOps. He started to describe the use-case of namespaces and its connection with patterns. What are our namespaces we can have when working with docker or kuberenetes, how we can control the abstraction and isolation in that way and how it will help us to have CAP ( CAP theorem ) which is, consistency, availability and partition tolerance. It was interesting to know that we can execute a hupervisor like KVM inside a container like Docker and it this way we can actually execute old Windows operating systems inside it :). He mentioned a Wormhole in docker which is caused a feature (!) that you can run docker in docker. It’s being done by mounting two following items and executing a loop command for sending commands to docker installed in host OS.

  • /usr/bin/docker
  • /usr/bin/docker.sock

Need to be fast to note next talks too so some of useful links about this talk are, kube-lego which is now recommended to switch to cert-manager. Then he showed a picture which there was something about jboss and I have no idea what that is and how it’s connected to tomcat! He mentioned a talk from OpenStack Summit where you can deploy OpenStack using k8s and helm ( I need to find the exact talk ). Replicaset of k8s and the comparison between stateful and stateless databases ( for example kafka is a stateful one ) TODO: check this out. Slides are here.

Okey … next one was done by Christian Daschner and Patrick Gruhn from Replex. quite a business talk I would say but some nice stuff were mentioned about the infrastructure and it’s provisioning and monitoring like: Terraform, rkt, Mesosphere, nmon, lsof, collectd, RackTables, Nagios, DataDog, Loggly, Sysdig, Heapster, influxDB. They were focusing much on the processing power we are wasting and how to recognize it and also how we can manage our server places in a rack. For future, it’s good to seach about Hybrid Cloud stuff.

Oh, wait! Need to find a power outlet! … Ok, now it’s getting charged 🙂 …

Next, we are going to have Roland Huß ( Twitter, Github, Blog, Leanpub ) from Red Hat talking about Kubernetes Patterns. After brief overview of the features in k8s, he is defining the desgin pattern and mentioning books of A Pattern Language and Design Patterns and Kubernetes Patterns. He talked about the patterns we have in different sections of k8s and gave examples that are mentioned here. He mentioned a blog post about writing software patterns and looks quite interesting to me as well as the blog itself by Martin Fowler who works at Thought Works. His slides are also here ( Backup ). After his talk it was time for the last one of this day, Container Software Supply Chain presented by Peter Rossbach ( Twitter ) and Niclas Mietz ( Twitter, Xing, Github ) which mostly gave some general and useful principals for managing docker containers. I tool some images and will share it here. They were using Harbor which is a private docker registry. This container registery is also interesting: Quay

Just to finish with, during the searches I had, I find this webpage, which might be interesting, need to check it out later.