CoreDNS Pod Replica Gets Continuously Scaled Down by the Deployment-Controller

Problem

  • One replica of the CoreDNS pod is getting scaled down continuously by the Deployment-Controller and the Kube-DNS__Autoscaler scaling up the replica at the same time.
Replica Set
Copy
  • Kube-DNS__Autoscaler keep scaling the replica back up to 2
Kube-DNSAutoscaler
Copy

Environment

  • Platform9 Managed Kubernetes - v5.6.4 and Higher.

Cause

  • This is as expected from an AutoScaler perspective because **preventSinglePointFailure** is default set to true.
Autoscaler default value
Copy
  • The Addon-Operator show the log that it was updating the CoreDNS-Addon continuously.
Addon operator log
Copy
  • Users may also see DNS resolution failures.

Resolution

  • This issue has been fixed in PMK v5.9.1. Upgrade the environment to v5.9.1+ to fix the issue.
Type to search, ESC to discard
Type to search, ESC to discard
Type to search, ESC to discard