Skip to content

Queue farm is a job queue that executes in order, backed by redis

License

Notifications You must be signed in to change notification settings

ekreative/queue-farm

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

41 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Queue farm

Queue farm is a job queue that executes in order, backed by redis.

Build Status

Creating jobs

Use the manager and push new jobs to it.

const queueFarm = require('queue-farm')

const manager = queueFarm.createManager()
manager.push('queue-name', {some: 'task'})
manager.push('another-queue-name', {some: 'task'})

Each named queue is executed in order, but different queues will be run out of order.

Running jobs

const queueFarm = require('queue-farm')

const worker = queueFarm.createWorker(async (queue, [job]) => {
  console.log(`Handle a job from ${queue}`, job)
})
worker.listen()

The worker will run indefinitely by default, call drain to run whilst there are still jobs visible.

API

queueFarm.createManager([options])

  • options <Object>
    • namespace <String> Default is 'queue-farm', used to run different instances on the same Redis server.
    • redis Either a Redis instance or connection string, uses 127.0.0.1:6379 by default. The Redis connection should exclusive to Queue Farm, make a new connection if you need to use Redis for other purposes.
    • deleteJobs <Bool> Default is true, when false the job hash isn't deleted from Redis, useful for debugging, but you need to manually clear up your Redis database .
    • timeout <Int> Default is 30000, default time (in milliseconds) to wait for a job to be completed before its counted as failed and should be retried.
    • maxAttempts, <Int> Default is 3, number of times the job will be attempted before it is discarded.
    • id, <String> Default is '{os.hostname()}:XXXX', where XXXX is 4 random digits. An id for the manager, useful to distinguish multiple running workers.
  • Returns: <queueFarm.Manager>.

Class queueFarm.Manager

Event: 'push'

  • queue <String> Name of the queue job.
  • jobId <String> Id of job pushed.
  • job * Data for the handler.
  • options <Object>.

New job pushed.

Event: 'delete'

  • queue <String> Name of the queue job.
  • jobId <String> Id of job pushed.
  • success <Bool>> true if the job was deleted.

Job deleted.

async manager.push(queue, job, [options])

  • queue <String> Name of the queue to push job to.
  • job * Data for the handler, will be JSON encoded for storage.
  • options <Object>
    • maxAttempts, <Int> Number of times the job will be attempted before it is discarded.
    • timeout, <Int> Time (in milliseconds) to wait for a job to be completed before its counted as failed and should be retried.
  • Returns: <String> The jobs id.

Put a new job in a queue.

async manager.del(queue, jobId)

  • queue <String> Name of the queue to delete the job from.
  • jobId <String> The id returned from push.
  • Returns: <Bool> true if the job was deleted, false if it was already performed.

Delete a job from the given queue.

queueFarm.createWorker([options], handler)

  • options <Object> Extends queueFarm.createManager options
    • noActiveQueuesSleep <Int> Default 30000. Time (in milliseconds) to sleep when there are no active queues.
    • allEmptySleep <Int> Default 1000. Time (in milliseconds) to sleep when all the queues are empty.
    • checkEmptyIterations <Int> Default is 50, how many iterations to wait before clearing active jobs queue.
    • concurrent <Int> Default is 1, how many jobs can be run concurrently.
    • batchSize <Int> Default is 1, how many jobs to return in a batch.
  • handler async <Function> Function that receives jobs to be processed. Args are (queue, jobs), just the same as passed to push. If errors are throw the job will be retried, up to retry limit.
  • Returns: <queueFarm.Worker>.

Class queueFarm.Worker extends queueFarm.Manager

Event: 'start'

  • queue <String> Name of the queue job is starting on.
  • jobIds [<String>] Ids of jobs starting now.
  • jobs [*] Datas of jobs starting now.

Started the job.

Event: 'finish'

  • queue <String> Name of the queue job is finished on.
  • jobIds [<String>] Ids of jobs finished now.
  • jobs [*] Datas of jobs starting now.

Finished handling job.

Event: 'handler-error'

  • queue <String> Name of the queue job has errored on.
  • jobIds [<String>] Ids of jobs that errored.
  • error * The error thrown by the handler.
  • jobs [*] Datas of jobs starting now.

Emitted when the handler throws an error.

Event: 'deleted-error'

  • queue <String> Name of the queue job has errored on.
  • jobIds [<String>] Ids of jobs that errored.
  • error Error An error object .

Emitted when trying to run a deleted job.

Event: 'retry-limit'

  • queue <String> Name of the queue job has errored on.
  • jobIds [<String>] Ids of jobs that errored.
  • datas [<String>] JSON encoded job datas.
  • error Error An error object.

Emitted when a job retry limit is ready.

Event: 'invalid-data'

  • queue <String> Name of the queue job has errored on.
  • jobIds [<String>] Ids of jobs that errored.
  • datas [<String>] JSON encoded job datas.
  • error Error An error object.

Emitted when trying to decode job data

Event: 'all-empty'

Emitted when all active queues are empty, before sleep.

Event: 'no-active-queues'

Emitted when there are no active queues, before sleep.

async worker.listen()

Start listening for jobs.

async worker.stop()

Stops listening for jobs. Prevents more jobs being taken off the queue, running jobs will finish.

async worker.drain

Listens for jobs until all the queues are empty and then returns.

queueFarm.createLogger(manager, logger, [options])

  • manager <queueFarm.Manager>
  • logger <Logger> | <Console>
  • options <Object> Extends queueFarm.createManager options
    • extended <Bool> Default false. To include the extended meta in the logs.
    • exceptEvents [<String>]Names of events not to listen for. Defaults to ['all-empty', 'no-active-queues']
    • defaultMeta * Data to be added to log meta.
  • Returns: <queueFarm.Logger>.

Class queueFarm.Logger

logger.on()

Start logging events.

logger.off()

Stop logging events.

How it works

Redis keys used:

Type Key Value Description
Set {namespace}:active queue names Used to track which queues exist.
Hash {namespace}:job:{jobId}.queuedAt timestamp Time job was queued.
Hash {namespace}:job:{jobId}.data string Time job was queued.
Hash {namespace}:job:{jobId}.maxAttempts int Max times to attempt this job.
Hash {namespace}:job:{jobId}.attempts int Times this job has been attempted.
Hash {namespace}:job:{jobId}.attemptAt timestamp Last time this job was attempted.
Hash {namespace}:job:{jobId}.timeout int Timeout (milliseconds) for this job.
Hash {namespace}:job:{jobId}.deleteAt timestamp Time job was deleted.
Hash {namespace}:job:{jobId}.finishAt timestamp Time job was finished.
Hash {namespace}:job:{jobId}.maxAttemptsAt timestamp Time job hit max attempts.
Hash {namespace}:job:{jobId}.errorAt timestamp Time job has error.
Hash {namespace}:job:{jobId}.dataDeletedErrorAt timestamp Time job has error because its data has been deleted since dequeuing.
Hash {namespace}:job:{jobId}.dataInvalidErrorAt timestamp Time job has error because its data couldnt be decode from JSON.
Hash {namespace}:job:{jobId}.workerId string Id of the worker handling this job.
Hash {namespace}:job:{jobId}.managerId string Id of the manager that pushed this job.
List {namespace}:queue:{queue} job ids List of jobs on the queue.
String {namespace}:active:{queue}:fetchAt timestamp Time job was taken from the queue, used to timeout the job.
List {namespace}:active:{queue} job ids List of active jobs for the queue.

Redis keys when adding a job:

  • Start MULTI
  • Add {queue} to {namespace}:active
  • Update {namespace}:job:{jobId}.queuedAt
  • Update {namespace}:job:{jobId}.data
  • Update {namespace}:job:{jobId}.maxAttempts
  • Update {namespace}:job:{jobId}.timeout
  • Update {namespace}:job:{jobId}.managerId
  • Add {jobId} to left of {namespace}:queue:{queue}
  • EXEC

When deleting a job:

  • Start MULTI
  • Delete {namespace}:job:{jobId} or update {namespace}:job:{jobId}.deleteAt
  • Delete {jobId} from {namespace}:queue:{queue}
  • EXEC

When taking a job from the queue:

  • Start a watch on {namespace}:active:{queue}
  • If there is a job in {namespace}:active:{queue}
    • Stop watching {namespace}:active:{queue}
    • Check if it is stalled and run it
  • Start MULTI - also fails if {namespace}:active:{queue} has changed
  • Get {jobId} from right of {namespace}:queue:{queue} and add {jobId} to left of {namespace}:active:{queue} (RPOPLPUSH)
  • Update {namespace}:active:{queue}:fetchAt
  • EXEC

Check if job has stalled:

  • Start a watch on {namespace}:active:{queue}:fetchAt
  • Get {namespace}:active:{queue}:fetchAt and {namespace}:job:{jobId}.timeout
  • If timeout is ok unwatch and do nothing
  • Use MULTI to update {namespace}:active:{queue}:fetchAt - so fails if already updated

Handling a job:

  • Get {namespace}:job:{jobId}
  • Checks attempts, if its >= maxAttempts the job is deleted
  • Start MULTI
  • Update {namespace}:job:{jobId}.attempts
  • Update {namespace}:job:{jobId}.attemptAt
  • Update {namespace}:job:{jobId}.workerId
  • EXEC
  • Start handler

Job error:

  • Start MULTI
  • Add {jobId} to right of {namespace}:queue:{queue}
  • Remove {jobId} from {namespace}:active:{queue}
  • Update {namespace}:job:{jobId}.errorAt
  • EXEC

Job complete:

  • Start MULTI
  • Delete {jobId} from {namespace}:active:{queue}
  • Delete {namespace}:job:{jobId} or Update {namespace}:job:{jobId}.finishAt
  • EXEC

Check for empty queue:

  • Watch {namespace}:active:{queue} and {namespace}:queue:{queue}
  • Get length of {namespace}:queue:{queue}
  • Get length of {namespace}:active:{queue}
  • If both are empty then start MULTI
  • Delete {queue} from {namespace}:active
  • Delete {namespace}:active:{queue}:fetchAt
  • EXEC

About

Queue farm is a job queue that executes in order, backed by redis

Resources

License

Stars

Watchers

Forks

Packages

No packages published