Kubernetes Contributor Summit North America 2024

The day-to-day, week-to-week, release-to-release job of being a maintainer of Kubernetes is hard. Perhaps unnecessarily so? In this session we can talk about what EXACTLY makes it hard, and what parts of that could be eliminated, streamlined, or delegated. We have lots of people who want to help, how can we enable them to? Where could we apply development energy to once-and-for-all overcome some of the "papercuts" that we all live with.

Discuss the future of hierarchical scheduling and possible improvements on k8s to better support dynamic batch workloads

Gateway API (SIG-NETWORK) introduced Policies as a declarative mechanism to enhance Kubernetes objects like Services and network extensions such as Gateways and HTTPRoutes. These Policies–more generally referred to as Metaresources–are the central piece of a pattern known as Policy Attachment, which is heading to new levels of maturity, evolving from initially an inventive way to expand solidified stable APIs, now into a new standard in the making across several networking-related implementations for contextual API extensions. This session will delve into the concept and evolution of Metaresources & Policy Attachment, exploring its goals, current state, and future potential.

Description: Begins with a short interest and feedback session to collect and socialize gaps in pod lifecycle management today. Followed by a working session to enumerate specific gaps in documentation and tests relating to the pod lifecycle, and brainstorm, including how to evaluate for feasibility, architectures to unify the state transition flow of pod lifecycle management (ref: state transition).
The output of this working session is

• a list of interested contribs
• a doc representing the current state, gaps in documentation and tests
• an appendix in the doc with brainstorming outputs from the working session on alternatives/updates to pod lifecycle management

Moderator: @lauralorenz (yes I am attending in person)

Description: Modern hardware changed our view on the native resources (CPU, memory). We want to talk on what and how it can be potentially improved on all levels: possible UX in Pod spec, per pod/container granularity to specify preferences, CRI API, relation to DRA, dynamic node resize, roles of kubelet and container runtimes.