Documentation
¶
Overview ¶
Package gcpscale holds the request-driven instance pool and injectable clock that Cloud Run style emulators share.
Index ¶
Constants ¶
const ( // DefaultIdleTimeout is how long an idle instance above the minimum is kept; Cloud Run documents "up to 15 minutes". DefaultIdleTimeout = 15 * time.Minute // DefaultMaxInstances is Cloud Run's default instance cap. DefaultMaxInstances = 100 )
Variables ¶
var ErrClosed = errors.New("instance pool is closed")
ErrClosed means the pool was shut down.
var ErrRetired = errors.New("instance was retired while starting")
ErrRetired means the instance was drained or removed before it finished starting.
var ErrSaturated = errors.New("no instance became available: the request timed out in the pending queue")
ErrSaturated means every instance slot stayed busy until the pending-queue timeout (HTTP 429 on Cloud Run).
Functions ¶
This section is empty.
Types ¶
type Clock ¶
type Clock interface {
Now() time.Time
NewTimer(d time.Duration) Timer
AfterFunc(d time.Duration, f func()) Timer
}
Clock abstracts time so idle scale-down and request queueing can be driven by tests.
type Config ¶
type Config[T any] struct { Start func(ctx context.Context) (T, error) Stop func(T) Limits func() Limits Idle func() time.Duration Clock Clock Base context.Context //nolint:containedctx // lifetime of the instances, set by the worker IdleTimeout time.Duration }
Config wires a Pool to the thing it scales. Start gets a context cancelled once its instance is retired (after Stop for a started one). Limits and Idle run under the pool lock and must not call back into the pool.
type Limits ¶
Limits are the scaling settings read afresh on every decision. Zero Max means DefaultMaxInstances, zero Concurrency means 1.
type Pool ¶
type Pool[T any] struct { // contains filtered or unexported fields }
Pool keeps instances of one revision: requests take a concurrency slot, new instances start while below the maximum, and idle ones above the minimum are removed.
func (*Pool[T]) Acquire ¶
Acquire reserves a request slot, starting an instance when all slots are taken and the pool is below its maximum, and queueing for max(10s, 3.5 x observed start time) otherwise (then ErrSaturated).
func (*Pool[T]) Close ¶
func (p *Pool[T]) Close()
Close stops every instance and rejects further requests. Starts in flight are cancelled and stopped as they finish.
func (*Pool[T]) Drain ¶
func (p *Pool[T]) Drain()
Drain retires every instance but keeps the pool open: later requests start fresh ones.
func (*Pool[T]) Ensure ¶
Ensure brings the pool to at least n instances (capped at the maximum) and waits until they are ready.
Source Files