Skip to main content

Scheduled AKS Scaling

Context​

K8SaaS provides dynamic auto-scaling feature with AKS. Based on the CPU and the memory requested by the pods, AKS adds or removes a nodes in the cluster. The auto-scaler works fine, but it could impact the customer experience waiting for a new node join the cluster.

When the usage is predictable, the scheduled AKS scaling works better.

How it works ? A job is scheduled when you want (every night, every hour etc...) and update the node pool size of your cluster.

For example:

  • node pool of 7 nodes during working hours
  • node pool of 2 nodes during night hours

Limitations:

  • the job resizes the cluster to the minimum of nodes (mandatory) before scaling up again

Use case​

  • Optimize costs
  • Provide better performances to endusers

What to do ?​

Ask for the enabling of this feature on the TrustNest K8SaaS Service catalog

Please precise:

  • the configuration of the scheduler
  • the min number of nodes
  • the max number of nodes

HOWTO​

Know the deployed scheduled AKS configuration ?​

Using azure portal:

  • Go to Azure Portal
  • Search for "Azure Kubernetes Services" in the search bar
  • Select your cluster in the list (if you do not see it, please add the subscription provided by email during the onboarding)
  • Click on Tags

You should see:

here

  • MinLowNode: gives you the minimum of nodes in the low mode
  • MaxLowNode: gives you the maximum of nodes in the low node
  • MinHighNode: gives you the minimum of nodes in the high mode
  • MaxHighNode: gives you the maximum of nodes in the high mode