Supporting Long-Lived Pods Using a Simple Kubernetes Webhook

Explore a two-sided system for supporting long-lived pods in Kubernetes through this conference talk. Learn how to implement an admission webhook that injects tolerations into pods, indicating their need for extended runtime. Discover the complementary node tainting system that marks nodes with their uptime. Understand how this approach enables scheduling long-lived pods on younger nodes, reducing the risk of early termination. Gain insights into writing a simple Kubernetes admission webhook, implementing the node tainting mechanism, and avoiding common pitfalls. Examine real-world metrics from Slack's implementation of this long-lived pod support system.

Supporting Long-Lived Pods Using a Simple Kubernetes Webhook - Clément Labbe, Slack