Struct CascadeDetector 

pub struct CascadeDetector { /* private fields */ }

Tracks failure propagation across DAG regions for cascade-aware routing.

A “region” is the set of tasks reachable from the “heaviest root” of each task — the root ancestor that is the source of the most downstream tasks. For tasks with multiple roots (diamond patterns), we pick the single root that covers the most descendants, preventing over-aggressive deprioritisation (C8 fix).

Implementations

impl CascadeDetector

pub fn new(config: CascadeConfig) -> CascadeDetector

Create a new detector.

pub fn record_outcome( &mut self, task_id: TaskId, succeeded: bool, graph: &TaskGraph, )

Record a task outcome and update the health of its primary region.

pub fn is_cascading(&self, task_id: TaskId, graph: &TaskGraph) -> bool

Returns true when the primary region of task_id is in cascade failure.

pub fn deprioritized_tasks(&self, graph: &TaskGraph) -> HashSet<TaskId>

Returns the set of task IDs that should be deprioritized due to cascade failure.

Returns an empty set when no region is cascading, avoiding unnecessary reordering.

pub fn reset(&mut self)

Reset all region health counters.

Called by DagScheduler::inject_tasks() because graph topology has fundamentally changed — old failure counts no longer reflect the new task set (C13 fix).

Trait Implementations

impl Debug for CascadeDetector

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.