node-gru

Gru

Gru is a node clustering helper, because he is the master of the minions

Is a Fork of https://github.com/hunterloftis/throng with quite a few improvements relating to having a separate startup and running phase.

A single master can control two kinds of workers:

• Generic workers. Workers who all perform the same task

• Dedicated workers. Workers who each perform a unique task.

Usage

import { gru } from 'gru'

const numCpus = os.cpus().length

gru({
    workers: process.env.NODE_ENV === 'production' ? numCpus : 1,
    lifetime: Infinity,
    master: async () => {
        // Master initialisation here
    },
    start: async () => {
        // Generic worker initialisation
    },
    dedicatedWorkers: {
        uniqueTask1: () => {
            // Dedicated worker initialization
        },
        uniqueTask2: () => {
            // Dedicated worker initialization
        },
    },
})

API

gru options

workers

Number of generic workers, if 0, starts in process (useful for debugging). Defaults to # of cpu cores

lifetime

ms to keep cluster alive or 'until-killed'

grace

ms grace period after worker SIGTERM (default 5000)

logger

Compatible with Bunyan, Winston and TypeScript-log style logging interfaces, logs gru activity

master

The master process callback.

Optionally can return a promise if there is startup work before worker is considered started.

If the returned promise resolves a value, it will be passed to the workers.

example

gru({
    master: async () => {
        const config = await loadConfig()

        return { config }
    },
    start: ({ masterArgs }) => {
        masterArgs.config // will be the config returned from the master process
    },
})

start

The worker callback.

If the worker returns a promise and it rejects, it signals to gru that the worker failed to start and will not be restarted. Once a worker has started the lifecycle policy will apply (ie workers which crash will be restarted)

dedicatedWorkers

The set of unique worker processes.

example

gru({
    master: async () => {
    },
    start: () => {
    },
    dedicatedWorkers: {
        express: () => {
            // Start an Express server
        },
        fileIngestion: () => {
            // Monitor a directory for zip files to process
        },
    },
})