Edgecase (kubernetes on the edge): afternoon summaries

Tags: kubernetes, python

(Some summaries of the May 2022 Dutch Edgecase meeting).

Edge: the new data center - Chris Urwin

Kubernetes is appealing on the edge. What every developer wants is a datacenter under their desk. With k8s (and especially k3s) they can.

  • Kubenetes helps a lot by giving you a real, complete environment to manage your workloads. No more custom scripts to monitor whether your docker still runs and restart it if needed. That’s all handled.

  • The challenge about kubernetes: it is big and bloated. The solution for this is k3s which is a full, certified kubernetes distribution but easy and quick to install and maintain.

Edge, what is it? To him, “anything outside of a data center” is edge.

  • Near edge: telco, media, communications. Bigger deploys. Often you see kubernetes installs being offered instead of regular applications.

  • Far edge: retail, banking, education, argriculture, utilities. Separate devices that need managing. K3S shines here (also in combination with “SLE micro”).

K3s is really open source. Build and born in the cloud. The rest of suse Rancher is too. Lightweight cloud-native edge stack, reliable and secure edge inffrastructure. They aim at low maintenance.

That “SLE micro”: OS build for the edge. Enterprise-grade, but especially for edge compute. A lightweight, immutable OS. Build for containers.

Provisioning a Civo private region for edge computing - Dinesh Majrekar

“Turtles all the way down”: they’re using kubernetes to provide kubernetes to customers. They’re trying to use only cloud stuff. They were the first managed k3s server provider. 90 second cluster boot time :-)

They try to be sustainable. In the UK they use biofuel for the emergency generators, for instance. They’re also doing things with “compute powered water heater” solutions: small computers for in individual homes to use exess heat to warm up water.

He showed what they’re doing regarding hardware. Mostly they deal with entire racks. The core is that they want to fully remotely manage such racks. From power-on to fully operational, they now manage it within six hours.

Inspiration usecase: deploying pods to watch pots in greenhouses - Gerrit Tamboer & Sander ter Schure

Dümmen Orange is a company that cross-breeds flowers. It is actually the biggest in the world. They do it in greenhouses world-wide. A greenhouse: a warm and humid environment. What they produce are seedlings for other companies to grow further.

Their main problem: people don’t scale very well. Finding more plant biologists is hard. They’re trying to do more with camera’s and other automatic sensors. The camera images are fed into an artificial intelligence deep learning system that monitor the growth and opening/closing of the flowers. Images taking terabytes of data, so processing in the greenhouse itself is an advantage: at the edge.

Now… how to get the client along on the journey towards kubernetes?

First they did a deployment of the deep learning application to a simple k3s cluster. The application previously took a lot of computing resources on their laptops. Now they could use the kubernetes version without their laptops running hot.

Second, they moved it over to the cloud. Azure in this case as they were already heavily integrated with all things microsoft. One of the things they started using was the azure key vault, as kubernetes’ own solution is kinda basic. And of course they used managed databases and the managed kubernetes service.

Third, they went fully live. In the greenhouses (so: the edge) the customer set up the hardware. Then they used ansible playbooks to deploy k3s, argocd and the deep learning application. A 3 node cluster, btw.

vanrees.org logo

About me

My name is Reinout van Rees and I work a lot with Python (programming language) and Django (website framework). I live in The Netherlands and I'm happily married to Annie van Rees-Kooiman.

Weblog feeds

Most of my website content is in my weblog. You can keep up to date by subscribing to the automatic feeds (for instance with Google reader):