KEP-0000: Pod Network Health API

keps/sig-network/0000-pod-network-health/README.md

@@ -0,0 +1,90 @@
+# Pod Network Health API
+
+## Summary
+Kubernetes currently lacks a native mechanism to represent basic
+pod-to-pod network health such as reachability and latency.
+This KEP proposes a Kubernetes-native API to express these signals
+in a standardized and extensible way.
+
+## Motivation
+Network issues are one of the most common causes of outages in Kubernetes.
+Today, operators rely on ad-hoc scripts, CNI-specific tools, or
+external observability systems to diagnose pod-to-pod connectivity issues.
+
+A standardized API enables:
+- Faster diagnosis of networking issues
+- Vendor-neutral observability
+- Better tooling integration
+
+## Goals
+- Define a Kubernetes-native abstraction for pod network health
+- Represent basic signals such as reachability and latency
+- Remain CNI-agnostic and implementation-neutral
+- Introduce the API as alpha behind a feature gate
+
+## Non-Goals
+- Deep packet inspection
+- Mandatory probing behavior
+- Automatic remediation
+- Replacing service meshes or observability platforms
+
+## User Stories
+
+### Cluster Operator
+As a cluster operator, I want to know whether two pods can communicate
+so that I can debug outages faster.
+
+### Platform Engineer
+As a platform engineer, I want a standard API to surface network health
+signals that can be consumed by monitoring systems.
+
+## Proposal
+Introduce an alpha Kubernetes API resource that represents observed
+network health between a source pod and a target pod.
+
+The API focuses on **representation**, not how data is collected.
+
+## API Design (High-Level)
+The API may include:
+- Source pod reference
+- Target pod reference
+- Reachability status
+- Optional latency metrics
+- Timestamp of last observation
+
+Exact fields will be refined during review.
+
+## Implementation Details
+- Introduced as alpha
+- Feature gated
+- No default probing required
+- Implementations may be controller-based, node-based, or vendor-provided
+
+## Alternatives Considered
+- CNI-specific tooling (not portable)
+- External observability systems (not Kubernetes-native)
+- CLI-only debugging tools (not programmatic)
+
+## Risks and Mitigations
+**API stability risk**  
+Mitigated by alpha status and feature gating.
+
+**Performance impact**  
+Mitigated by avoiding mandatory probing.
+
+## Graduation Criteria
+
+### Alpha
+- API introduced behind feature gate
+- Experimental usage
+
+### Beta
+- Feedback from users
+- Stable semantics
+
+### GA
+- Production usage
+- Documented best practices
+
+## References
+- SIG-Network discussions (TBD)

keps/sig-network/0000-pod-network-health/kep.yaml

@@ -0,0 +1,9 @@
+title: Pod Network Health API
+kep-number: 0000
+authors:
+  - Sahichowdary
+owning-sig: sig-network
+participating-sigs:
+  - sig-node
+status: provisional
+creation-date: 2025-01-24