How I reduced pod startup times from minutes to seconds with intelligent image preloading

If Red Hat trusts OpenShift to run the control plane for their largest infrastructure orchestrator, the same pattern should apply to your smallest.

Recently, in our daily operations, we took a deep dive into the inner workings of linkerd-destination, one of the most critical components of the Linkerd control plane.

If you've ever wanted to develop a command line client for a Kubernetes API, especially if you've considered making your client usable as a kubectl plugin, you might have wondered how to make your client feel familiar to users of kubectl. In fact, the Kubernetes project provides two libraries to help you handle kubectl-style command line arguments in Go programs: clientcmd and cli-runtime (which uses clientcmd). This article will show how to use the former.

I'm starting a series of blog posts to explore CloudNativePG (CNPG), a Kubernetes operator for PostgreSQL that automates high availability in containerized environments.

OpenKruise Agents provides best practices for managing AI agent workloads in Kubernetes. It is a sub-project of the open-source workload project OpenKruise under the Cloud Native Computing Foundation (CNCF), specifically tailored for the AI agent domain. OpenKruise Agents accelerates AI agent deployment and makes it easily accessible to both AI algorithm scientists and infrastructure engineers.

In Amazon Web Services, each instance type has a different upper limit on how many Pods it can run. For example, an m5.large instance can only run 29 Pods, but an m5.4xlarge can run up to 234. The reason is that each EC2 instance can only have a limited number of IP addresses assigned to it.

In Kubernetes clusters, GPUs are expensive. Really expensive. And when your data science team is fighting over GPU resources like it’s Black Friday at Best Buy, you know something’s broken.

That was the challenge I faced a few years ago. I implemented this solution back then, and though it’s been a while, I thought now would be a great time to share the experience.

We had a k8s cluster with 8 NVIDIA A100 GPUs, and our data scientists were… unhappy. Small model training jobs waited hours for GPU access while large distributed training jobs hogged entire GPUs at 30% utilization.

The culprit? GPU allocation granularity. Kubernetes treats GPUs as atomic resources, you either get a whole GPU or nothing. There’s no middle ground.

I needed to fix this. And the solution wasn’t what I expected.

The Kubernetes ecosystem is undergoing a significant transformation in how it handles networking, shifting from traditional Ingress API and annotations to the more advanced Gateway API.

Cluster API (CAPI) is an open-source Kubernetes sub-project. Its goal is to bring Kubernetes-style, declarative APIs and controllers to the problem of bootstrapping, configuring, upgrading and operating entire Kubernetes clusters, treating clusters themselves as first-class Kubernetes resources rather than as external, manually-provisioned infrastructure.