Deep dive into the stable trafficDistribution API in Kubernetes 1.35, covering PreferSameNode and PreferSameZone options for reducing latency and cross-zone data transfer costs. Read on CloudKeeper →
Identified recurring patterns in customer support tickets at CloudKeeper and built automation around them. Created EKS review questionnaires, upgrade scripts, DQI tracking sheets, and standardized customer-facing documentation. Impact: Reduced time spent on common EKS issues Standardized customer review and upgrade processes Built reusable scripts for multi-version EKS upgrades Created internal KB articles to reduce repeated questions Tools: Bash, Python, Kubernetes, EKS
Explaining the EKS Provisioned Control Plane option, when it makes sense to use it, and when the standard control plane is enough. Read on CloudKeeper →
Led cost optimization engagements across 50+ customer environments at CloudKeeper. Worked on right-sizing pods, tuning VPA/HPA configurations, implementing Karpenter with optimal bin-packing, and moving eligible workloads to spot instances. Key results: Reduced one customer’s monthly bill by $20K on a $111K/month account (18% savings) Saved a logistics platform $15K+/month through EKS and Karpenter optimization Standardized cost review questionnaires and savings tracking across the team Tools: EKS, Karpenter, VPA, HPA, KEDA, Helm, Terraform ...
Diagnosing the common scenario where EKS pods stay in Pending state even when the VPC has available IP addresses. Covers ENI limits, prefix delegation, and subnet sizing. Read on CloudKeeper →
EKS upgrades across multiple Kubernetes versions are complex, especially in production environments with custom networking, service meshes, and stateful workloads. Built comprehensive upgrade tooling and documentation that was used across multiple customer engagements. What I built: Step-by-step upgrade scripts with pre-flight checks Rollback procedures for each upgrade stage Documentation covering add-on compatibility, API deprecations, and breaking changes Deep dives into Kubernetes 1.34 and 1.35 feature changes Tools: EKS, Kubernetes, Bash, Terraform ...
Walking through a complex EKS upgrade that resolved persistent production stability issues for a customer running critical workloads. Read on CloudKeeper →
Karpenter efficiently provisions nodes to match pod requests, but if developers request 4 CPUs for a pod that uses 0.2 CPUs, Karpenter will buy a node to fit that bloated request. It bin-packs waste, but it cannot fix the waste itself. Built an internal scheduling component that optimizes at the scheduling stage itself, not just at provisioning. Ran 14 internal iteration cycles to find the optimal bin-packing strategy that accounts for actual resource utilization, not just requested resources. ...
Took a monolithic PHP-based supercar auction platform and broke it into containerized services running on EKS. The platform included Moodle for training modules and Nginx as the web layer. What I did: Containerized 40+ microservices from bare-metal/VM setup Deployed on EKS with proper scaling, health checks, and resource limits Set up CI/CD pipeline for automated builds and deployments Configured monitoring and alerting Tools: EKS, Docker, Terraform, GitHub Actions, Nginx, PHP, Moodle ...