Hi all 👋 I’m working on FLINK-39176: Node Health Management & Node Quarantine Framework in Apache Flink, aiming to provide a unified mechanism for node-level quarantine and health management, with slot-allocation integration and operational visibility. The feature is split into 6 small, independently reviewable PRs: PR-1 – Introduce NodeHealthManager Abstraction https://github.com/apache/flink/pull/27701 Introduce

NodeHealthManager

interface Provide

Default

and

NoOp

implementations • Integrated into

ResourceManager

(no behavior change, backward compatible) PR-2 – Integrate NodeHealthManager with Slot Filtering https://github.com/apache/flink/pull/27711 Integrate health check into

FineGrainedSlotManager

Skip quarantined TaskManagers during slot allocation • Backward compatible via

NoOpNodeHealthManager

PR-3 – Add REST API for Node Quarantine https://github.com/apache/flink/pull/27712 • Add REST endpoints: POST quarantine DELETE quarantine GET list • Extend

ResourceManagerGateway

PR-4 – Configuration Support for Node Quarantine https://github.com/apache/flink/pull/27714 • Add

ManagementOptions

config:

node-quarantine.enabled

node-quarantine.default-duration

node-quarantine.max-duration

• Rename “management blocklist” → “node quarantine” (avoid confusion with FLIP-224) PR-5 – Expiration Cleanup Scheduler https://github.com/apache/flink/pull/27715 • Add scheduled expiration cleanup for quarantined nodes • Prevent stale entries accumulation PR-6 – Web UI Node Health Page https://github.com/apache/flink/pull/27716 • Add Web UI page for node quarantine visibility • Operational observability improvement This framework ensures: • No slot allocation on unhealthy/quarantined nodes • Clear separation from FLIP-224 speculative blocklist • Fully configurable and operationally manageable Would really appreciate reviews, especially on: • ResourceManager integration • Slot filtering semantics • REST API design consistency • Config naming and compatibility Thanks in advance 🙏