JA EN
shiichan

EKS Auto Mode Now Steers Around Broken Zones Automatically!

AWS What's New ★★★★ View Markdown
#aws#infrastructure#devops

Hi everyone, it's Shii-chan! Today I've got news for anyone running Kubernetes that's quietly awesome. EKS Auto Mode just got smarter about handling zone outages!

AWS What's New aws.amazon.com

What was announced?

This comes from AWS What's New. If you're running Amazon EKS Auto Mode, your clusters can now use Amazon Application Recovery Controller (ARC) zonal shift and autoshift.

ARC helps you manage and coordinate recovery across AWS Regions and Availability Zones (AZs), and EKS Auto Mode is the part of EKS that manages your compute for you. Now the two work together.

The story so far

Many of you spread applications across multiple AZs for availability. But when an AZ actually had trouble, shifting traffic away from it took real setup: flags to enable, permissions to grant, and Karpenter versions to keep track of.

What changes

If you're on EKS Auto Mode, you just enable ARC zonal shift on your cluster. No flags, no extra permissions, no Karpenter version management. And it costs nothing extra. Getting ready for AZ failures just got a lot easier.

Dive Deep

When a zonal shift kicks in, EKS Auto Mode automatically:

  • Stops provisioning new capacity in the impaired AZ
  • Halts voluntary disruptions, like consolidation and drift, for nodes in that zone
  • Also prevents voluntary disruptions in healthy zones if they depend on scheduling pods into the impaired zone

When the shift expires or gets canceled, normal operations resume automatically.

There are two ways to trigger a shift: start one manually, or let AWS manage it for you with autoshift. Autoshift also supports practice runs, so you can verify your cluster still works fine with one less AZ before it actually matters.

This is available in every AWS Region where EKS Auto Mode is offered.

Wrap-up

  • EKS Auto Mode now supports ARC zonal shift and autoshift
  • Enabling it automatically blocks new provisioning and disruptions in an impaired AZ
  • No flags, permissions, or Karpenter version management required
  • Practice runs let you validate cluster behavior ahead of time
  • No extra cost, available everywhere EKS Auto Mode is offered

If you run EKS across multiple AZs and want less manual work when things go wrong, this update is for you.

Related posts