We are starting a new series at Kubelist this week: Kubernetes Resources and Autoscaling. We are going to be taking a little break from aggregating new content and curate some lists based on things we think are important to know for Kubernetes. (If you're really hungry for the newest stuff check out: https://kube.news/ or @kubeweekly).
This week we bring you Resource Management from the basics to advanced. These articles answer: How are resources managed in Kubernetes? What is utilization? What is isolation? How do my pods get scheduled? Why do my pods get killed? And the essential wisdom of Kubernetes resource management: Always set resource and limits.
A classic case made for the use of a dynamic resource allocator within Kubernetes and the complications therein. Why should you have to set container resources manually? Checkout the reference implementation at: https://github.com/openshift-demos/resorcerer
Isolation, Right sizing, Utilization: The core components of kubernetes resource management. A stellar talk on the dangers of full utilization, stability over performance and the insights brought from Borg.
Technically it’s OpenShift but it mostly applies to Vanilla Kubernetes. This series is really stellar because it goes into a lot of depth about how resources are set and managed by Kubernetes. The series could be called: autoscaling from first principles.
ALWAYS set limits and cpu, got it? Readable intro to how kubernetes manages resources - how apps get scheduled and how they get killed.
BYOR - Why resource classes will(do) enable bring your own resources and what the community is doing to make it happen.
Some common theme in Kubernetes resource management: Cost creeps in, Over provisioning creeps in, Performance degradation creeps in. Learn how to avoid the creep with resource management primitives.
If you want to know how the kubernetes layer of resource management translates down to the docker layer which translates down to the kernel read here.
We end this issue with testing the k8s resources “limits” with load tests and telling us the results.
Wanna know how much resources you are using? Even at the cost level? Checkout out Purser!