limits

Documentation

Overview

Package limits enforces the runtime safety rails that bound one client's or one database's share of the server: a per-principal request rate limit (token bucket) and a per-database concurrency cap (so one hot database can't starve the others). The result caps and statement/transaction timeouts live in the engine and session store; this package is the admission control in front of them.

Index

• type Limiter
  • func New(perPrincipalRPS float64, maxConcurrentPerDB int) *Limiter
  • func (l *Limiter) Allow(principal, db string) (release func(), ok bool, reason string)
  • func (l *Limiter) InFlight(db string) int

Types

type Limiter

type Limiter struct {
	// contains filtered or unexported fields
}

Limiter admits or rejects a request before it reaches the engine. The zero value (via New with non-positive limits) admits everything, so limits are opt-in.

func New

func New(perPrincipalRPS float64, maxConcurrentPerDB int) *Limiter

New builds a Limiter. perPrincipalRPS<=0 disables rate limiting; maxConcurrentPerDB<=0 disables the concurrency cap.

func (*Limiter) Allow

func (l *Limiter) Allow(principal, db string) (release func(), ok bool, reason string)

Allow reserves capacity for one request by principal against db. On success it returns a release func that MUST be called when the request finishes (it frees the per-db concurrency slot); ok=false means the request is rejected — reason is "rate" (per-principal rate exceeded) or "busy" (per-db concurrency cap reached), so the caller can map it to 429 vs 503.

func (*Limiter) InFlight

func (l *Limiter) InFlight(db string) int

InFlight reports the current in-flight request count for db (for introspection / tests).