Kubernetes IP or ENI exhaustion runbook (summary)

A runbook in this repository, Kubernetes/runbooks/insufficient-ip-or-eni/README.md, for Pods that stay Pending with ENI or IP scheduling events. It uses Tencent Kubernetes Engine (TKE) VPC-CNI as its worked example, with every environment value as a placeholder.¹

Takeaways

• A Pending Pod is a broken link in a chain (node health, Pod-subnet IPs, node ENI/IP allocatable, IPAM and admission components); find the link before changing anything. See IP or ENI exhaustion runbook.
• Remediate in dependency order and restart workloads only when needed. See Safe change procedure.
• Close the incident only against explicit acceptance criteria.¹ See Incident closure criteria.

Footnotes

1. Kubernetes IP or ENI Exhaustion Scheduling Failure Runbook, original on GitHub ↩ ↩²