[FOLIO-2200] Implement K8 auto-scaling on EKS cluster Created: 31/Jul/19  Updated: 03/Jun/20  Resolved: 12/Aug/19

Status: Closed
Project: FOLIO
Components: None
Affects versions: None
Fix versions: None

Type: Task Priority: P3
Reporter: John Malconian Assignee: John Malconian
Resolution: Done Votes: 0
Labels: platform-backlog
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original estimate: Not Specified

Issue links:
Relates
relates to UXPROD-1827 CI-integrated continuous deployment (... Closed
Sprint: CP: sprint 69
Story Points: 5
Development Team: Core: Platform

 Description   

There are two types of auto-scaling concepts typically associated with Kubernetes. Horizontal auto-scaling is the concept of scaling pods based on system utilization. It is accomplished by configuring K8 resources based on observed application utilization metrics (namely CPU) that specifies how many replicas of a pod the replication controller should deploy. Node auto-scaling is the concept of scaling nodes up or down based on the capacity to deploy pods. Node autoscaling on EKS is implemented by integrating Kubernetes with AWS Auto-Scaling Groups so that the number of nodes will automatically increase or decrease based on pod capacity. There is no auto-scaling of Kubernetes nodes or pods enabled by default. We will need implement both in order to support FOLIO performance testing in a Kubernetes cluster as well as ensure we have enough node capacity to continuously deploy new modules.

Horizonal auto-scaling is described further here: https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale/

Node auto-scaling is described here: https://github.com/kubernetes/autoscaler/blob/master/cluster-autoscaler/cloudprovider/aws/README.md



 Comments   
Comment by John Malconian [ 31/Jul/19 ]

Installed the Kubenetes Metrics Server according to these instructions - https://docs.aws.amazon.com/eks/latest/userguide/metrics-server.html - as it is a requirement for horizontal auto-scaling. Installing the Metrics Server Helm chart via Rancher does not work since it requires installation in the kube-system namespace and Rancher does not seem to support that as an option.

Generated at Thu Feb 08 23:18:57 UTC 2024 using Jira 1001.0.0-SNAPSHOT#100246-sha1:7a5c50119eb0633d306e14180817ddef5e80c75d.