Cloud-native notes on Kubernetes, platform engineering, and modern infrastructure.
An engineering-oriented comparison of KAI-Scheduler’s Reservation Pod approach and HAMi’s hard isolation path, including trade-offs, failure modes (noisy neighbor), and how the two layers can complement each other.
An engineering-oriented guide to hetGPU: how a compiler + runtime stack can make one GPU binary run across NVIDIA/AMD/Intel/Tenstorrent, including SIMT vs MIMD, memory model gaps, and live kernel migration.
A practical boundary guide: Docker packages and runs containers, Kubernetes orchestrates and keeps services stable at scale, and OpenStack turns datacenter hardware into an IaaS resource pool (VM/network/storage).
A deep dive into gpu-manager startup, device interception, topology awareness, and allocation mechanics for Kubernetes GPU virtualization.
A structured walkthrough of CGroup concepts, V1/V2 differences, controllers, and hands-on troubleshooting.
Understand calling conventions, stack frames, call/ret behavior, debugging observation, and security implications from the assembly view.
Use structure, examples, and tools to connect ELF types, layout, relocations, and dynamic linking.
Combine Deployment rollingUpdate settings with PodDisruptionBudgets to keep availability during upgrades and node maintenance.
A structured workflow for diagnosing Pending pods, CrashLoopBackOff, traffic failures, and node-level issues—without guessing.
How CPU/memory requests and limits actually affect scheduling, throttling, OOMKills, and autoscaling.
A production-friendly approach to ServiceAccounts, Roles, and bindings that minimizes blast radius without breaking workflows.
Avoid crash loops and bad rollouts by using the right probe for the right job.
A step-by-step approach to introducing NetworkPolicy without breaking everything on day one.
Safely inspect a live Pod without baking debugging tools into production images.
How to make autoscaling predictable: right requests, sane HPA behavior, VPA recommendations, and capacity-aware cluster scaling.
Tell Kubernetes when an app is ready or needs a restart.
Deploy MySQL replication quickly using Helm charts.
Expose a Pod or Service to a local port for quick debugging.
Use replication for read scaling and recovery.
Provide stable DNS for StatefulSets without load balancing.
Let PVCs trigger storage provisioning automatically.
Provide stable identity and storage for stateful workloads.
Ephemeral volumes live with the Pod and fit cache or temp files.
Decouple storage providers and storage consumers.
Decouple configuration and sensitive data from images.
Learn how Pods share and persist data with volumes.
Run a single-instance MySQL with PVC, Deployment, and Service.
Stateful services need stable identity, storage, and ordered startup.
Manage resources with YAML and a reviewable change workflow.
Validate a new version with small traffic before scaling up.