Upgrades keep slipping
The cluster is several versions behind and nobody wants to be the one who breaks production.
Kubernetes consulting
EKS and AKS upgrades, builds and cost, done properly
Get your clusters onto a supported version, set up new ones the right way, and stop paying for node capacity your pods never use.
Hands-on, every day, in production
- AWS
- Azure
- Amazon EKS
- Azure AKS
- Kubernetes
- Terraform
- GitHub Actions
- Prometheus
- Grafana
- OpenTelemetry
Common Kubernetes problems we fix
Nodes half empty
Requests are guesses, so the cluster autoscaler adds nodes for capacity nobody uses.
Built by hand
The cluster was clicked together in the console and nobody can rebuild it with confidence.
What we do
Version upgrades
EKS and AKS, one minor version at a time.
- Deprecated API scan and manifest fixes
- Add-on compatibility (CNI, CoreDNS, ingress, CSI drivers)
- Blue/green node pool rollout
- Rollback plan for every step
New cluster builds
Repeatable, from code.
- Terraform modules for EKS or AKS
- Private networking and IAM / Entra ID integration
- Ingress, cert-manager, external-dns, secrets
- GitOps with Argo CD or Flux
Autoscaling and cost
Pay for what pods actually use.
- Requests and limits from real usage
- Karpenter or Cluster Autoscaler tuning
- Spot node pools for suitable workloads
- Graviton / arm64 node groups on EKS
Reliability
Fewer pages at night.
- Pod disruption budgets and probes
- Resource quotas and limit ranges
- Prometheus and Grafana for cluster health
- Runbooks your team can follow
An upgrade runbook, not a leap of faith
Every upgrade follows the same documented path, so it is boring in the best way:
- Scan for APIs removed in the target version and fix the manifests and Helm charts that use them.
- Check every add-on against the target version and upgrade the ones that need it first.
- Upgrade staging, run your smoke tests, and note anything unexpected.
- Upgrade the production control plane in an agreed window.
- Create new node pools on the new version, drain the old ones gradually, then remove them.
- Write down what changed, so the next upgrade takes half the time.
How we work
20-minute call
Current versions, add-ons, how you deploy and what is blocking the upgrade or the build.
Read-only assessment
Cluster configuration, deprecated APIs, add-on compatibility, node pools and cost.
Written plan and fixed quote
Step-by-step runbook with rollback points, timing and a fixed price.
Delivery with your team
Changes in agreed windows, staging first, documented so your team can run the next one.
Read-only access. NDA on request. No change is made without your written approval.
What it costs
Every engagement starts with a free 20-minute review, so you know the size of the opportunity before you spend anything.
Step 1
Free cloud bill review
Free
A 20-minute call and a one-page note naming the 3 biggest leaks in your bill.
- You send a Cost Explorer or Azure Cost Management export
- 20-minute call to understand your workload
- One-page note: 3 biggest leaks, with rough monthly value
- No access to your account needed
Step 3
Savings implementation
Fixed quote
I make the changes from the audit, with your approval on each one.
- Rightsizing and idle-resource cleanup
- Savings Plans, Reserved Instances and Azure reservations
- Storage tiering, gp2 → gp3, Graviton and Spot where safe
- Tagging, budgets and cost alerts
Fixed quote, or 20–25% of the first 6 months of verified savings.
Book a callStep 4
Managed DevOps / FinOps retainer
₹25,000 – ₹75,000 / month$600 – $1,800 / month
Ongoing care so the bill stays down and the platform stays current.
- Monthly cost review and report
- CI/CD and Kubernetes care, version upgrades
- On-call during business hours (IST)
- Direct line to the engineer, not a ticket queue
Questions people ask
How long does an EKS or AKS version upgrade take?
One minor version on a well-kept cluster is usually a few days of work including staging. Clusters several versions behind need one step per minor version, plus add-on and API changes, so plan for longer. The assessment gives you a firm estimate.
Can you upgrade without downtime?
In most cases, yes: the control plane is upgraded in place and node pools are replaced with new ones while workloads are drained gradually. That depends on workloads having more than one replica, pod disruption budgets and health checks, which the assessment checks first.
Why does an old Kubernetes version cost more?
Both EKS and AKS charge extra for clusters on versions that have left standard support (EKS extended support, AKS long-term support). Staying current avoids that charge as well as the security risk.
Do you set up new clusters too?
Yes: EKS or AKS built with Terraform, private networking, node pools, autoscaling (Karpenter or Cluster Autoscaler), ingress, certificates, secrets, GitOps deployment and monitoring, handed over with documentation.
Can you help with private AKS clusters for regulated workloads?
Yes. Private API server, private endpoints, egress through a firewall, Entra ID integration and Azure Policy are all part of a typical private AKS setup.
Talk through your cluster
Tell me your current version and what is blocking you. You will get a reply within one working day and a link to book a 20-minute call.
Or email hello@nextopslab.in