Struct BudgetChecker 

pub struct BudgetChecker { /* private fields */ }

Cross-partition budget checker with per-cycle caching.

Reads budget usage/limits once per scan cycle (not per candidate). This is MANDATORY for performance — without it, 50K blocked executions would produce 50K budget reads per cycle.

Implementations

impl BudgetChecker

pub fn new(config: PartitionConfig) -> Self

pub async fn check_budget( &mut self, client: &Client, budget_id: &str, ) -> Result<&BudgetCheckResult, SchedulerError>

Check a budget by ID. Reads from Valkey on first call per budget, caches for subsequent candidates in the same cycle.

Fail-closed: transport errors propagate as SchedulerError::ValkeyContext rather than being cached as Ok. Caching an Err would be wrong — a transient blip would then pin every candidate in this cycle to the same denial. Successful results (including HardBreach) are still cached so 50K blocked candidates sharing a budget do one read per cycle, not 50K.

pub fn reset(&mut self)

Clear the cache at the start of a new scheduler cycle.