Hey folks! I'm curious about your strategies for scaling down to zero during non-business hours. Do you have systems in place, like cron jobs or KEDA, to handle this? What areas do you manage? For example, are you scaling down the entire test cluster or just certain namespaces? Also, I'd love to hear about the Kubernetes flavor you're using—especially if it's ARO (Azure Red Hat OpenShift). Is it common to completely remove nodes during these hours? What challenges have you faced, and have you seen any significant cost savings? Thanks!
5 Answers
Just unplug everything! 😂 Though, on a serious note, there’s also autoscaling, combining horizontal and vertical scaling. My clusters typically drop to one small instance for core services when they aren't needed.
I don't scale down to zero, but I get quite close. With tools like the cluster autoscaler or Karpenter, it’s really straightforward to minimize resource usage.
We use EKS on AWS and shut down our testing environments over the weekend. Initially, we employed scheduled Lambda functions but later switched to Step Functions. One handles the shutdown, while another restarts the services. We also scale down our worker nodes and databases this way.
It’s interesting to consider this. If scaling production down to zero is necessary after hours, maybe your pricing needs adjusting. It’s all about finding balance.
KEDA can indeed scale your workloads down to zero based on specific metrics or cron schedules. Just remember, you still pay for the nodes, not the containers. After addressing workload scaling, ensure your nodes can downscale themselves. I use Karpenter for underutilized nodes, which can go down to zero. Spot instances are great for cost savings during business hours too!
Related Questions
Can't Load PhpMyadmin On After Server Update
Redirect www to non-www in Apache Conf
How To Check If Your SSL Cert Is SHA 1
Windows TrackPad Gestures