Skip to content
Agencei

Expertise

Kubernetes expertise: orchestrating your containers in production

Kubernetes is the standard container orchestration platform. It handles workload placement, scaling, restarts after failures, networking between services, configuration, secrets and progressive rollouts, all declaratively, from manifests stored in version control.

Agencei deploys and operates managed Kubernetes clusters (mainly Amazon EKS) or lightweight ones (k3s) for Java, Node.js, Python applications and CMSs. We cover Kubernetes manifests and Helm charts, GitOps with Argo CD, ingress, certificates, observability and workload security.

This expertise is for organisations that run several services, have availability and scaling requirements, deploy frequently, or have inherited a poorly understood cluster that is hard to maintain, upgrade and secure.

Use cases

Microservices platform

Deploying dozens of services with discovery, centralised configuration, resource limits and network policies.

Zero-downtime deployments

Rolling updates, health probes, blue-green or canary deployments through Argo Rollouts to ship several times a day safely.

Autoscaling

Horizontal Pod Autoscaler on CPU, memory or business metrics, and node autoscaling with Karpenter or Cluster Autoscaler.

GitOps

The cluster state is described in Git; Argo CD or Flux applies changes, making every deployment traceable and reversible.

Observability

Prometheus, Grafana, Loki and OpenTelemetry for metrics, logs and traces, with alerts on service-level indicators.

Taking over and hardening an existing cluster

Audit of versions, RBAC permissions, resources without limits and deployment practices, followed by a remediation plan.

Project types

  • EKS cluster setup
  • Docker Compose to Kubernetes migration
  • GitOps pipeline with Argo CD
  • Helm charts for your applications
  • Observability and alerting
  • Cluster audit and hardening
  • k3s cluster for modest environments

Integration with the rest of the stack

  • Kubernetes + AWS (EKS)

    Managed cluster, autoscaled nodes, ALB load balancing through the Load Balancer Controller, EBS and EFS storage, per-pod IAM identities.

  • Kubernetes + Docker + CI/CD

    Pipelines build the images, update the manifests in Git and let Argo CD synchronise the cluster.

  • Kubernetes + Spring Boot or Node.js

    Liveness and readiness probes, graceful shutdown, configuration through ConfigMaps and Secrets, and resource limits aligned with the JVM or Node.js.

  • Kubernetes + PostgreSQL

    Managed database outside the cluster (RDS) by default; the CloudNativePG operator when the database must run inside the cluster.

  • Kubernetes + ingress and certificates

    NGINX or Traefik ingress, cert-manager for Let's Encrypt certificates and External DNS for automatic records.

Examples of problems solved

Pods stuck in CrashLoopBackOff

Log and event analysis, fixing overly aggressive probes, missing variables or dependencies unavailable at startup.

Pods evicted or killed for exceeding memory

Defining realistic requests and limits, tuning the JVM or Node.js according to those limits, and monitoring actual consumption.

Deployments causing 502 errors

Introducing readiness probes, graceful shutdown with a delay, PodDisruptionBudgets and a suitable rolling update strategy.

Cluster that cannot be upgraded

Inventory of deprecated APIs, chart and operator upgrades, step-by-step version bumps tested on a staging cluster.

Cluster cost out of control

Node consolidation, autoscaling with Karpenter, Spot instances for tolerant workloads and right-sizing of requests.

Frequently asked questions

Do we really need Kubernetes?

Not always. For one or two applications, Docker Compose or ECS Fargate is often enough. Kubernetes becomes relevant with several services, fine-grained scaling needs or a team that deploys very frequently.

Managed or self-hosted cluster?

Managed (EKS, GKE, AKS) in the vast majority of cases: the control plane is maintained by the provider. Self-hosting or k3s is justified by cost, isolation or specific environment constraints.

What is GitOps?

A practice where the desired cluster state is stored in Git and applied automatically by a tool such as Argo CD. Every change is reviewed, traced and reversible with a simple Git revert.

How do you secure a Kubernetes cluster?

Least-privilege RBAC, network policies, non-root containers, scanned images, externalised secrets, regular upgrades and auditing of control plane access.

Can you train our team?

Yes. We support teams during setup, document operating procedures and transfer skills so the cluster remains under control internally.

Tell us about your project

Describe your need in a few lines: we come back to you with a first analysis and the next steps.