We, at Cybercom, have been working with OpenShift for years now. In last two years we have grown our OpenShift expertise to about ten people. Some of us had opportunity to attend OpenShift Commons Gathering Helsinki, and the following RedHat Forum Helsinki on the next day.
OpenShift Commons Gathering Helsinki 2018 was the first Commons Gathering in Helsinki, and before that we have only had two OpenShift meetups in Finland. Here are few takeaways from the gathering, and links in the end.
Renaming from Origin to OKD
The upstream version of OpenShift was renamed from Origin to OKD in release 3.10. That's no change under the hood, but major change in branding. Diana Mueller explained that the renaming was made to emphasize the Kubernetes technology underneath. In short OKD is a Kubernetes distribution. The old name Origin dates back to versions before OpenShift v3, that based on completely different technology.
Upcoming OpenShift v4 and Core OS
The upcoming OpenShift v4 was mentioned. It seems that the new version essentially means merging of Core OS technologies to OpenShift. There seems to be lot's of synergy in Core OS acquisition, but especially three features were highlighted.
- Atomic Host will be replaced with minimal Red Hat Core OS. That should make upgrade of host operating system easier.
- Red Hat Quay registry, that will have more enterprise features.
- OpenShift Operator Framework. There was a lot's of talk about operators in the gathering.
In Short, Operator is CoreOS term for Kubernetes Custom Resource Definition + Controller. Custom Resource Definition adds new resources to Kuberenes API and the Controller is the code that manages the new resource type. In a simple and generic example Operator is used to provision a database, like postgresql, with simple "give me a resource" command.
There is also the older Service Catalog concept, that was introduced in 3.7 release, and it provides similar functionality. One big difference is that Kubernetes Controllers are limited to manage Kubernetes resources, but Service Catalog Brokers are able to provision the resource in external services, like RDS.
However, there is more in the Operator concept. There is idea of replacing human labor also in the lifecycle phase. It is called Operator Lifecycle Management, and it should keep the managed resources up to date. This is an interesting concept to be aware of.
Also, it was good to see that monitoring improvements were on the roadmap, because there is a need for better visibility in OpenShift cluster. We have been using prometheus for monitoring OpenShift, and it is great to see that the official installation is getting more mature.
Open Practice Library
Jeremy Brown introduced Red Hat's Open Innovation Labs concept. They support customers by teaching the Open Innovative approach to take advantage of new technology. That was inspiring and there were lot's of good links and resources in the talk. Especially, there was a note about Open Practices Library, that is collection on methods used in the Open Innovation Labs context.
OpenShift Cloud Functions
Serverless has lot's of hype nowadays, and it was great to see that there is almost ready solution for running serverless on top of OpenShift, and that is called OpenShift Cloud Functions. In practice it is Open Whisk running on OpenShift.
Especially interesting was Tero Ahonen's comment about Knative compatibility. Knative is another serverless implementation on Kubernetes project, but it is backed by Google. Compatibility and portability of applications on serverless frameworks is important, as the open serverless frameworks are evolving. Few years back, we didn't know which container orchestration system would get the largest adoption. There were Mesos, Nomad, Swarm, and now we only have Kubernetes left. Similarly, there is fragmentation in serverless tools and we will find out the best solution in few years.
Pains of the on-premise installation
Three use cases were presented, which were all interesting. I would like to especially mention Elisa story, as it resonates with our experiences when helping with the customers in on-premise OpenShift installations. Antti Seppälä presented a story, how their number of Kubernetes installation on top of OpenStack evolved to managed multi-datacenter OpenShift cluster. We, at Cybercom, have also been involved in OpenStack for multiple years, and we found the similar lesson that OpenStack is not the answer for developers' generic automation need, and OpenShift/Kubernetes are better for that.
Also, the story, that you have to
- add intranet routers,
- manage routing to external services and
- architecture for high availability,
is so common. The reference architecture is not complete enough for most of the needs and it would be great to have these common needs addressed in the documentation more clearly.
Generally, there were surprisingly many new git repositories and new resources to study. I had a false feeling that I know most of the resources. Thanks to participants for sharing all this. Note, some of the following are not from the gathering, but they support the discussion.
- Information about CoreOS acquisition https://www.redhat.com/en/about/press-releases/red-hat-unveils-roadmap-coreos-integration-red-hat-openshift
- Open Practices Library https://openpracticelibrary.com/
- Hands on learning https://learn.openshift.com
- Collection of OpenShift installation practices http://v1.uncontained.io
- Open Innovation Labs https://github.com/rht-labs
- A yet another code for managing the OpenShift Cluster https://github.com/redhat-cop/casl-ansible
- Knative vs OpenShift Cloud Functions https://blog.openshift.com/state-of-serverless-in-kubernetes-knative-and-openshift-cloud-functions/
Waiting for Red Hat's recap blog. Thank you for the event.