Upgrade to Pro — share decks privately, control downloads, hide ads and more …

KEDA - Mastering Scale to Zero by Daniel Hasselwander

KEDA - Mastering Scale to Zero by Daniel Hasselwander

Slides from the following meetup: https://www.meetup.com/de-DE/microsoft-azure-zurich-user-group/events/292938328/

KEDA is an Event Driven Autoscaler for workloads running on Kubernetes. It provides mechanisms to scale those workloads depending on external systems, such as messaging or telemetry systems.

This talk will provide an introduction to KEDA and the underlying concepts. The usage of KEDA will be demonstrated on Azure with an AKS cluster running a demo application, which is autoscaled through KEDA. We will see how to setup KEDA and configure it to scale the application depending on the amount of messages in an Azure Service Bus Topic

About Daniel:
I have been working as an Azure Engineer at ti&m in Zurich since 2021. Ti&m is an IT service provider offering services along the entire IT value chain. My role is to support customers in the selection of Azure Cloud components and to implement them afterwards. My main technologies are container technologies (e.g. AKS) and Terraform as an IaaC tool.


More Decks by Azure Zurich User Group

Other Decks in Technology


  1. What is Keda ? − Kubernetes-based Event Driven Autoscaler −

    Works alongside standard components − Installable with helm − Doku on keda.sh
  2. Agent – Scale to Zero activate and deactivate Deployments «keda-operator»

    container How it works ? KEDA Parts Metrics Server exposes event data for HPA to scale out «keda-operator-metrics- apiserver» container
  3. ScaledObjects Events Deployment etc. How it works ? Custom Ressources

    (CRD) ScaledJobs Events Kubernetes Job TriggerAuthentication Namespace useable credentials ClusterTriggerAuthentication Cluster useable credentials
  4. KEDA – HTTP Scaler − An HTTP "event" is not

    well defined − no out-of-the-box single system that can provide an API to measure the current number of incoming HTTP events or requests. − must to be integrated into the HTTP routing system
  5. KEDA – HTTP Scaler Add-On − KEDA core project has

    purposely not built generic HTTP-based scaling − Beta Status − Try it and give the community feedback − GitHub - kedacore/http-add-on: Add-on for KEDA to scale HTTP workloads
  6. KEDA – When to use it − Can be your

    default scale mechanism − Scale-To-Zero != Realtime − Azure Function in your Cluster
  7. KEDA – Good to know − Short polling only for

    activate deployment − Don´t combine ScaleObject with HPA − Can use multiple scaler − Scale on your own systems − Azure CLI implement a KEDA extension for Functions