NFD: Simplifying Cluster Administration through Automated Node Labels, Taints, and Annotations
- Track: HPC, Big Data & Data Science devroom
- Room: UA2.118 (Henriot)
- Day: Saturday
- Start: 17:30
- End: 18:00
- Video only: ua2118
- Chat: Join the conversation!
Kubernetes is being used in more diverse use cases that require specialized hardware and system configurations per node. The Node Feature Discovery (NFD) project was initially created as a simple node labeler daemon to tackle the difficulties of heterogeneous clusters. However, recently, NFD has acquired numerous new capabilities, from custom node labeling to extended resources, node tainting, annotations, and topology discovery.
NFD has a broad range of built-in capabilities to detect hardware and system configurations and has various extension points for third-party plugins. Additionally, NFD provides a CRD-based rule system that is expressive and flexible for updating node properties according to its discovered capabilities.
This talk will demonstrate how NFD can be utilized to simplify cluster administration by automating aspects of node management, such as labeling and tainting.
Speakers
Carlos Eduardo Arango Gutierrez |