What Kubernetes support includes
You can take a single item, for example just a version upgrade, or hand over the whole cluster.
Cluster deployment
I build a cluster that is described in code and can be recreated from scratch.
- ✓managed Kubernetes in the cloud
- ✓on-premise: K3s or kubeadm on your servers
- ✓ingress, certificates, storage
- ✓Terraform and Helm instead of manual commands
Version upgrades
Kubernetes ships a new minor version roughly every 4 months. I upgrade on a plan, not when the provider forces it.
- ✓deprecated API checks before upgrading
- ✓control plane, nodes and add-ons
- ✓one node at a time, no downtime
- ✓a rollback plan
Resources and stability
Most pod crashes come from wrong limits and missing health checks.
- ✓requests and limits based on real usage
- ✓liveness and readiness probes
- ✓pod and node autoscaling
- ✓PodDisruptionBudget for safe upgrades
Monitoring and logs
You see not only what broke but why.
- ✓cluster, node and application metrics
- ✓logs from all pods in one place
- ✓Telegram alerts
- ✓dashboards for the team
CI/CD and GitOps
A release should come from Git and roll back with one command, not go out via kubectl from a laptop.
- ✓Helm charts for services
- ✓GitLab CI or GitHub Actions pipelines
- ✓GitOps with ArgoCD
- ✓development and production environments
Security
I close the typical holes left in clusters after a quick start.
- ✓RBAC and separate access for people and CI
- ✓network policies between services
- ✓secrets kept out of Git
- ✓image vulnerability scanning
Backups
The cluster can be recreated from code, the data can't. So backups and restore tests are a must.
- ✓manifests and persistent volumes
- ✓databases backed up outside the cluster
- ✓storage in S3
- ✓scheduled restore tests
Cluster audit
I review a cluster inherited from a previous team and write a plan.
- ✓versions and support dates
- ✓resources and node costs
- ✓access, secrets, network policies
- ✓an answer to whether you need Kubernetes at all
Kubernetes support pricing
Fixed rates. Work outside the support plan is estimated and agreed in advance.
| Format | Price | What's included |
|---|---|---|
| Cluster audit | from $230 | versions, resources, access, deployment, monitoring, backups; a report with a plan |
| Support | $250/month | watching alerts and nodes, planning upgrades, incident response |
| One-off tasks | $35/hour | version upgrades, monitoring, moving services into the cluster |
Example estimates
| Task | Estimate |
|---|---|
| Upgrade the cluster by one minor version | 2-6 hours |
| Monitoring and alerts for the cluster and applications | 4-8 hours |
| Move services from Docker Compose into the cluster | from 10 hours |
| Set up deployment with ArgoCD | 4-8 hours |
How we work
- 1
Read-only access
I start with a read-only kubeconfig: I can look at the cluster without breaking anything.
- 2
Cluster map
Versions, nodes, load, what is deployed and how, where secrets and data are kept.
- 3
Risks
I note what will break first: deprecated APIs, pods without limits, missing backups.
- 4
Work plan
I propose the order of work. Sometimes that means moving to K3s or Docker Compose instead of fixing the cluster.
- 5
Support
If needed, I stay on support: scheduled upgrades, alerts, helping the team with deployments.
Common problems
The cluster was inherited from a previous DevOps
No documentation, the version is long out of support, and nobody knows what an upgrade will break.
Pods restart and nobody looks into it
OOMKilled, CrashLoopBackOff, overloaded nodes. Usually it's limits and probes, not "Kubernetes being flaky".
The cluster costs more than it should
Nodes at 15% utilization, test environments running around the clock, more load balancers than services.
Manual releases
Releases go out via kubectl apply from a laptop, and nobody can repeat yesterday's deployment.
Do you need Kubernetes?
Kubernetes solves real problems, but it costs money and time to run. A quick check:
| It fits if | Probably not needed if |
|---|---|
| more than 5-10 services and several environments | 1-3 services on one or two servers |
| the load varies and you need autoscaling | the load is steady and predictable |
| zero-downtime releases matter | a short outage during a release is fine |
| someone on the team will work with it | nobody can look after a cluster |
At MomsLab, moving from full K8s to K3s cut the infrastructure budget by 40%.
Kubernetes, K3s or Docker Compose
Three options small teams usually choose between:
| Kubernetes | K3s | Docker Compose | |
|---|---|---|---|
| For | many services, growth, cloud | the same on 1-5 nodes with fewer resources | a few services on one server |
| Complexity | high | medium | low |
| Compatibility | the standard | same API, same Helm charts | its own format, no cluster |
| Ecosystem | everything | almost everything | Docker only |
Moving between Kubernetes and K3s is nearly seamless: the manifests and charts stay the same. Going from Compose to Kubernetes means describing the services again as manifests or Helm charts.
Technologies
- Clusters
- Managed KubernetesK3skubeadmAWS EKS
- Tools
- kubectlk9s
- Deployment
- HelmArgoCDGitLab CIGitHub Actions
- Monitoring
- PrometheusGrafanaLokiVictoriaMetrics
- Security
- RBACNetworkPolicyTrivyVault
- Infrastructure
- TerraformAnsibleAWSHetzner
Case studies
App for moms: from K8s to K3s
A small project ran on full Kubernetes, and the infrastructure cost more than its scale required.
Moved the services to K3s with the same Helm charts and GitLab pipelines, set up a fault-tolerant entry point with DNS failover, added metrics, logs and backups with restore tests.
Autonomous tractors: a cluster for IoT
An autonomous tractor fleet needed a control platform, real-time video from the machines over GSM and a secure network in the field.
Deployed an AWS EKS cluster with ArgoCD releases and isolated the tractor network with WireGuard and mutual TLS authentication.
Why work with me
Certified administrator
CKA from The Linux Foundation: the exam is taken on a live cluster, not as a quiz.
I don't sell Kubernetes
If K3s or Docker Compose is enough for your project, I'll say so in the audit.
Pay for work, not for headcount
Supporting one cluster shouldn't cost as much as an engineer's salary.
Everything in Git
The cluster and service configuration lives in your code. Any engineer can pick it up.
Frequently asked questions
How is Kubernetes different from Docker? +
Docker packages an application into a container and runs it on one server. Kubernetes manages containers across many servers: restarts failed ones, spreads the load, updates without downtime. The images stay the same.
Managed Kubernetes or on-premise? +
Small teams are usually better off with a managed service: the provider runs the control plane. An on-premise cluster on K3s or kubeadm is cheaper on dedicated servers and needed when data must stay inside your network, but it takes more care.
How often does Kubernetes need upgrading? +
A new minor version comes out roughly every 4 months, and each is supported for about 14 months. Without upgrades a cluster falls out of support within a year, and cloud providers start upgrading it forcibly.
How much does Kubernetes support cost? +
Support is $250 a month: alerts, node health, upgrade planning and incident response. Work beyond that is billed at $35/hour. A cluster audit starts at $230.
What is K3s and will it work for us? +
K3s is a lightweight Kubernetes with the same API: the same manifests and Helm charts, but lower resource requirements and simpler upkeep. It fits 1-5 nodes and a limited budget well.
What access do you need? +
A read-only kubeconfig is enough for an audit. For actual work I need cluster admin access, which can be granted temporarily and revoked.
Send a request
Tell me where the cluster runs, how many nodes and services it has and what worries you. I'll say where to start and whether you need Kubernetes at all.