Struct ResourceMonitor 

pub struct ResourceMonitor<T: Float + Debug + Send + Sync + 'static> { /* private fields */ }

Resource monitor for tracking and managing system resources

Implementations

impl<T> ResourceMonitor<T>

where T: From<f64> + Display + Float + Debug + Send + Sync + 'static,

pub fn new(constraints: ResourceConstraints<T>) -> Self

Create a new resource monitor

pub fn start_monitoring(&mut self) -> Result<()>

Start resource monitoring

pub fn stop_monitoring(&mut self) -> Result<()>

Stop resource monitoring

pub fn update_usage(&self) -> Result<()>

Sample every tracker and record a snapshot (F12).

This is called once per generation from NeuralArchitectureSearch::check_resource_constraints; before that wiring it was never called at all, so current_usage stayed at ResourceUsage::default() forever and check_resource_violations could never fire.

The snapshot’s derived fields are all measured or absent: active_processes, memory_pressure, cpu_load_average and gpu_utilization used to be hard-coded to 1 / Low / 0.6 / 0.8.

The clock is read once and reused for the timestamp and the retention cutoff, so a snapshot cannot be pruned by an Instant::now() taken microseconds later.

pub fn check_violations(&self) -> Result<Vec<ResourceViolation<T>>>

Check for resource violations

pub fn optimize_resources(&self) -> Result<Vec<OptimizationAction<T>>>

Optimize resource usage

pub fn get_current_usage(&self) -> ResourceUsage<T>

Get current resource usage.

Recovers from a poisoned mutex instead of panicking (F22): the guarded data is a plain snapshot with no invariants that a panicking writer could have broken, so one panic elsewhere must not make this getter panic forever.

pub fn get_usage_history(&self) -> Vec<ResourceSnapshot<T>>

Get usage history. Poison-recovering, for the same reason as ResourceMonitor::get_current_usage.

pub fn get_monitoring_state(&self) -> MonitoringState

Get monitoring state

pub fn set_trackers(&mut self, monitors: Vec<Box<dyn ResourceTracker<T>>>)

Replace the resource trackers. This is the injection point for real telemetry: hand in a SystemResourceTracker::with_telemetry built over a source that can measure what the default cannot (F12).

pub fn set_alert_handlers(&mut self, handlers: Vec<Box<dyn AlertHandler<T>>>)

Replace the alert handlers.

pub fn set_optimization_strategies( &mut self, strategies: Vec<Box<dyn ResourceOptimizer<T>>>, )

Replace the optimization strategies.

pub fn telemetry_source_names(&self) -> Vec<&str>

Names of the telemetry sources currently backing this monitor.

pub fn update_constraints(&mut self, constraints: ResourceConstraints<T>)

Update constraints

pub fn generate_report(&self) -> ResourceReport<T>

Generate resource report

pub fn check_resource_violations(&self) -> Result<Vec<ResourceViolation<T>>>

Check for resource constraint violations

pub fn get_usage_summary(&self) -> ResourceUsageSummary<T>

Get a summary of resource usage

Trait Implementations

impl<T: Float + Debug + Send + Sync + 'static> Debug for ResourceMonitor<T>

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

Formats the value using the given formatter.