← Back to Blog

EKS upgrades without downtime: the checklist I use every time

Upgrading Kubernetes sounds simple… until your application goes down in production.

I’ve seen teams rush upgrades and end up with broken deployments, failed pods, or unexpected downtime.

Over time, I built a simple checklist that I follow every time I upgrade an Amazon EKS cluster — and it has saved me multiple times.

Why EKS upgrades can break things

When you upgrade EKS, you're not just upgrading Kubernetes.

• API versions may change
• Node groups get replaced
• Pods restart

If your app is not prepared, users will feel it.

Step 1: Check compatibility first

Before touching anything, I always check:

• Kubernetes version compatibility
• Deprecated APIs
• Helm charts / manifests

One small API change can break your entire deployment.

Step 2: Make sure your app can handle restarts

This is where most issues happen.

I always verify:

• Readiness probes are properly configured
• Liveness probes are not too aggressive
• Multiple replicas are running (minimum 2)

If you have only one pod, downtime is guaranteed.

Step 3: Configure rolling updates

Your deployment strategy matters a lot.

strategy:
  type: RollingUpdate
  rollingUpdate:
    maxUnavailable: 0
    maxSurge: 1

This ensures: - No downtime - New pods start before old ones stop

Step 4: Upgrade node groups safely

I never upgrade all nodes at once.

Instead:

• Create a new node group
• Drain old nodes gradually
• Monitor pods during migration

This avoids sudden failures.

Step 5: Watch everything (very important)

During upgrade, I continuously monitor:

• Pod status
• Error rates
• Latency

Even small spikes can indicate problems.

Step 6: Have a rollback plan

Never upgrade without a rollback plan.

I always keep:

• Previous node group ready
• Old deployment configs

If something breaks, you should recover fast.

What this approach solves

After following this checklist:

• No downtime during upgrades
• Fewer surprises
• More confidence in production changes

Final thoughts

EKS upgrades are not risky — if you prepare properly.

Most problems come from skipping basics like readiness checks and rolling updates.

Keep it simple. Test first. Monitor everything.

---

Written by Adarsh Singh — DevOps Engineer