Bulkhead

A bulkhead limits the calls running at the same time. It protects a dependency from too many concurrent calls, and protects your process from piling up work on a slow dependency.

import { Bulkhead, Tenace } from '@julr/tenace'

const bulkhead = new Bulkhead({ concurrency: 10, queue: 100, maxWait: '1s' })

const policy = Tenace.policy().withTimeout('2s').with(bulkhead)

Share the instance to share the limit between policies.

API

policy.withBulkhead(bulkhead: Bulkhead | BulkheadOptions)
new Bulkhead(options: BulkheadOptions)

Option	Type	Default	Description
concurrency	number	required	Calls running at the same time. An integer of 1 or more, or Infinity.
queue	number	0	Calls waiting for a slot. An integer of 0 or more, or Infinity.
maxWait	Duration	no limit	Longest time a call waits in the queue before it is rejected.
name	string	'bulkhead'	Name in events and in describe().

An invalid concurrency or queue throws a RangeError when the bulkhead is built.

A Bulkhead instance exposes concurrency, running (the calls running now), and queued (the calls waiting now).

Behavior

When a slot is free, the call runs. When every slot is busy, the call waits in the queue if there is room, and fails with BulkheadFullError otherwise. With the default queue: 0, a call fails as soon as every slot is busy.

Queued calls run in arrival order. A queued call fails with BulkheadFullError when it waits longer than maxWait, and leaves the queue when its signal aborts. A call whose signal aborts at the moment it gets its slot does not run.

Error channel

Adds BulkheadFullError. Its running and queued fields hold the counts when the call was rejected.

Events

Event	Fields	When
bulkhead.rejected	bulkhead, running, queued	A call is rejected.