Desmond  Gerber

Desmond Gerber

1646642520

Agenda: MongoDB-backed Job Scheduling

A light-weight job scheduling library for Node.js

Agenda offers

  • Minimal overhead. Agenda aims to keep its code base small.
  • Mongo backed persistence layer.
  • Promises based API.
  • Scheduling with configurable priority, concurrency, repeating and persistence of job results.
  • Scheduling via cron or human readable syntax.
  • Event backed job queue that you can hook into.
  • Agenda-rest: optional standalone REST API.
  • Inversify-agenda - Some utilities for the development of agenda workers with Inversify.
  • Agendash: optional standalone web-interface.

Feature Comparison

Since there are a few job queue solutions, here a table comparing them to help you use the one that better suits your needs.

Agenda is great if you need a MongoDB job scheduler, but try Bree if you need something simpler (built by a previous maintainer).

FeatureBullBeeAgenda
Backendredisredismongo
Priorities 
Concurrency
Delayed jobs 
Global events  
Rate Limiter  
Pause/Resume  
Sandboxed worker  
Repeatable jobs 
Atomic ops 
Persistence
UI 
REST API  
Optimized forJobs / MessagesMessagesJobs

Kudos for making the comparison chart goes to Bull maintainers.

Installation

Notice

In order to support new MongoDB 5.0 and mongodb node.js driver/package the next release (5.x.x) of Agenda will be major. The required node version will become >=12. The mongodb dependency version will become >=3.2.

Install via NPM

npm install agenda

You will also need a working Mongo database (v3) to point it to.

CJS / Module Imports

for regular javascript code, just use the default entrypoint

const Agenda = require("agenda");

For Typescript, Webpack or other module imports, use agenda/es entrypoint: e.g.

import { Agenda } from "agenda/es";

NOTE: If you're migrating from @types/agenda you also should change imports to agenda/es. Instead of import Agenda from 'agenda' use import Agenda from 'agenda/es'.

Example Usage

const mongoConnectionString = "mongodb://127.0.0.1/agenda";

const agenda = new Agenda({ db: { address: mongoConnectionString } });

// Or override the default collection name:
// const agenda = new Agenda({db: {address: mongoConnectionString, collection: 'jobCollectionName'}});

// or pass additional connection options:
// const agenda = new Agenda({db: {address: mongoConnectionString, collection: 'jobCollectionName', options: {ssl: true}}});

// or pass in an existing mongodb-native MongoClient instance
// const agenda = new Agenda({mongo: myMongoClient});

agenda.define("delete old users", async (job) => {
  await User.remove({ lastLogIn: { $lt: twoDaysAgo } });
});

(async function () {
  // IIFE to give access to async/await
  await agenda.start();

  await agenda.every("3 minutes", "delete old users");

  // Alternatively, you could also do:
  await agenda.every("*/3 * * * *", "delete old users");
})();
agenda.define(
  "send email report",
  { priority: "high", concurrency: 10 },
  async (job) => {
    const { to } = job.attrs.data;
    await emailClient.send({
      to,
      from: "example@example.com",
      subject: "Email Report",
      body: "...",
    });
  }
);

(async function () {
  await agenda.start();
  await agenda.schedule("in 20 minutes", "send email report", {
    to: "admin@example.com",
  });
})();
(async function () {
  const weeklyReport = agenda.create("send email report", {
    to: "example@example.com",
  });
  await agenda.start();
  await weeklyReport.repeatEvery("1 week").save();
})();

Full documentation

Agenda's basic control structure is an instance of an agenda. Agenda's are mapped to a database collection and load the jobs from within.

Table of Contents

Configuring an agenda

All configuration methods are chainable, meaning you can do something like:

const agenda = new Agenda();
agenda
  .database(...)
  .processEvery('3 minutes')
  ...;

Agenda uses Human Interval for specifying the intervals. It supports the following units:

seconds, minutes, hours, days,weeks, months -- assumes 30 days, years -- assumes 365 days

More sophisticated examples

agenda.processEvery("one minute");
agenda.processEvery("1.5 minutes");
agenda.processEvery("3 days and 4 hours");
agenda.processEvery("3 days, 4 hours and 36 seconds");

database(url, [collectionName])

Specifies the database at the url specified. If no collection name is given, agendaJobs is used.

agenda.database("localhost:27017/agenda-test", "agendaJobs");

You can also specify it during instantiation.

const agenda = new Agenda({
  db: { address: "localhost:27017/agenda-test", collection: "agendaJobs" },
});

Agenda will emit a ready event (see Agenda Events) when properly connected to the database. It is safe to call agenda.start() without waiting for this event, as this is handled internally. If you're using the db options, or call database, then you may still need to listen for ready before saving jobs.

mongo(dbInstance)

Use an existing mongodb-native MongoClient/Db instance. This can help consolidate connections to a database. You can instead use .database to have agenda handle connecting for you.

You can also specify it during instantiation:

const agenda = new Agenda({ mongo: mongoClientInstance.db("agenda-test") });

Note that MongoClient.connect() returns a mongoClientInstance since node-mongodb-native 3.0.0, while it used to return a dbInstance that could then be directly passed to agenda.

name(name)

Sets the lastModifiedBy field to name in the jobs collection. Useful if you have multiple job processors (agendas) and want to see which job queue last ran the job.

agenda.name(os.hostname + "-" + process.pid);

You can also specify it during instantiation

const agenda = new Agenda({ name: "test queue" });

processEvery(interval)

Takes a string interval which can be either a traditional javascript number, or a string such as 3 minutes

Specifies the frequency at which agenda will query the database looking for jobs that need to be processed. Agenda internally uses setTimeout to guarantee that jobs run at (close to ~3ms) the right time.

Decreasing the frequency will result in fewer database queries, but more jobs being stored in memory.

Also worth noting is that if the job queue is shutdown, any jobs stored in memory that haven't run will still be locked, meaning that you may have to wait for the lock to expire. By default it is '5 seconds'.

agenda.processEvery("1 minute");

You can also specify it during instantiation

const agenda = new Agenda({ processEvery: "30 seconds" });

maxConcurrency(number)

Takes a number which specifies the max number of jobs that can be running at any given moment. By default it is 20.

agenda.maxConcurrency(20);

You can also specify it during instantiation

const agenda = new Agenda({ maxConcurrency: 20 });

defaultConcurrency(number)

Takes a number which specifies the default number of a specific job that can be running at any given moment. By default it is 5.

agenda.defaultConcurrency(5);

You can also specify it during instantiation

const agenda = new Agenda({ defaultConcurrency: 5 });

lockLimit(number)

Takes a number which specifies the max number jobs that can be locked at any given moment. By default it is 0 for no max.

agenda.lockLimit(0);

You can also specify it during instantiation

const agenda = new Agenda({ lockLimit: 0 });

defaultLockLimit(number)

Takes a number which specifies the default number of a specific job that can be locked at any given moment. By default it is 0 for no max.

agenda.defaultLockLimit(0);

You can also specify it during instantiation

const agenda = new Agenda({ defaultLockLimit: 0 });

defaultLockLifetime(number)

Takes a number which specifies the default lock lifetime in milliseconds. By default it is 10 minutes. This can be overridden by specifying the lockLifetime option to a defined job.

A job will unlock if it is finished (ie. the returned Promise resolves/rejects or done is specified in the params and done() is called) before the lockLifetime. The lock is useful if the job crashes or times out.

agenda.defaultLockLifetime(10000);

You can also specify it during instantiation

const agenda = new Agenda({ defaultLockLifetime: 10000 });

sort(query)

Takes a query which specifies the sort query to be used for finding and locking the next job.

By default it is { nextRunAt: 1, priority: -1 }, which obeys a first in first out approach, with respect to priority.

Agenda Events

An instance of an agenda will emit the following events:

  • ready - called when Agenda mongo connection is successfully opened and indices created. If you're passing agenda an existing connection, you shouldn't need to listen for this, as agenda.start() will not resolve until indices have been created. If you're using the db options, or call database, then you may still need to listen for the ready event before saving jobs. agenda.start() will still wait for the connection to be opened.
  • error - called when Agenda mongo connection process has thrown an error
await agenda.start();

Defining Job Processors

Before you can use a job, you must define its processing behavior.

define(jobName, [options], handler)

Defines a job with the name of jobName. When a job of jobName gets run, it will be passed to handler(job, done). To maintain asynchronous behavior, you may either provide a Promise-returning function in handler or provide done as a second parameter to handler. If done is specified in the function signature, you must call done() when you are processing the job. If your function is synchronous or returns a Promise, you may omit done from the signature.

options is an optional argument which can overwrite the defaults. It can take the following:

  • concurrency: number maximum number of that job that can be running at once (per instance of agenda)
  • lockLimit: number maximum number of that job that can be locked at once (per instance of agenda)
  • lockLifetime: number interval in ms of how long the job stays locked for (see multiple job processors for more info). A job will automatically unlock once a returned promise resolves/rejects (or if done is specified in the signature and done() is called).
  • priority: (lowest|low|normal|high|highest|number) specifies the priority of the job. Higher priority jobs will run first. See the priority mapping below
  • shouldSaveResult: boolean flag that specifies whether the result of the job should also be stored in the database. Defaults to false

Priority mapping:

{
  highest: 20,
  high: 10,
  normal: 0,
  low: -10,
  lowest: -20
}

Async Job:

agenda.define("some long running job", async (job) => {
  const data = await doSomelengthyTask();
  await formatThatData(data);
  await sendThatData(data);
});

Async Job (using done):

agenda.define("some long running job", (job, done) => {
  doSomelengthyTask((data) => {
    formatThatData(data);
    sendThatData(data);
    done();
  });
});

Sync Job:

agenda.define("say hello", (job) => {
  console.log("Hello!");
});

define() acts like an assignment: if define(jobName, ...) is called multiple times (e.g. every time your script starts), the definition in the last call will overwrite the previous one. Thus, if you define the jobName only once in your code, it's safe for that call to execute multiple times.

Creating Jobs

every(interval, name, [data], [options])

Runs job name at the given interval. Optionally, data and options can be passed in. Every creates a job of type single, which means that it will only create one job in the database, even if that line is run multiple times. This lets you put it in a file that may get run multiple times, such as webserver.js which may reboot from time to time.

interval can be a human-readable format String, a cron format String, or a Number.

data is an optional argument that will be passed to the processing function under job.attrs.data.

options is an optional argument that will be passed to job.repeatEvery. In order to use this argument, data must also be specified.

Returns the job.

agenda.define("printAnalyticsReport", async (job) => {
  const users = await User.doSomethingReallyIntensive();
  processUserData(users);
  console.log("I print a report!");
});

agenda.every("15 minutes", "printAnalyticsReport");

Optionally, name could be array of job names, which is convenient for scheduling different jobs for same interval.

agenda.every("15 minutes", [
  "printAnalyticsReport",
  "sendNotifications",
  "updateUserRecords",
]);

In this case, every returns array of jobs.

schedule(when, name, [data])

Schedules a job to run name once at a given time. when can be a Date or a String such as tomorrow at 5pm.

data is an optional argument that will be passed to the processing function under job.attrs.data.

Returns the job.

agenda.schedule("tomorrow at noon", "printAnalyticsReport", { userCount: 100 });

Optionally, name could be array of job names, similar to the every method.

agenda.schedule("tomorrow at noon", [
  "printAnalyticsReport",
  "sendNotifications",
  "updateUserRecords",
]);

In this case, schedule returns array of jobs.

now(name, [data])

Schedules a job to run name once immediately.

data is an optional argument that will be passed to the processing function under job.attrs.data.

Returns the job.

agenda.now("do the hokey pokey");

create(jobName, data)

Returns an instance of a jobName with data. This does NOT save the job in the database. See below to learn how to manually work with jobs.

const job = agenda.create("printAnalyticsReport", { userCount: 100 });
await job.save();
console.log("Job successfully saved");

Managing Jobs

jobs(mongodb-native query, mongodb-native sort, mongodb-native limit, mongodb-native skip)

Lets you query (then sort, limit and skip the result) all of the jobs in the agenda job's database. These are full mongodb-native find, sort, limit and skip commands. See mongodb-native's documentation for details.

const jobs = await agenda.jobs(
  { name: "printAnalyticsReport" },
  { data: -1 },
  3,
  1
);
// Work with jobs (see below)

cancel(mongodb-native query)

Cancels any jobs matching the passed mongodb-native query, and removes them from the database. Returns a Promise resolving to the number of cancelled jobs, or rejecting on error.

const numRemoved = await agenda.cancel({ name: "printAnalyticsReport" });

This functionality can also be achieved by first retrieving all the jobs from the database using agenda.jobs(), looping through the resulting array and calling job.remove() on each. It is however preferable to use agenda.cancel() for this use case, as this ensures the operation is atomic.

disable(mongodb-native query)

Disables any jobs matching the passed mongodb-native query, preventing any matching jobs from being run by the Job Processor.

const numDisabled = await agenda.disable({ name: "pollExternalService" });

Similar to agenda.cancel(), this functionality can be acheived with a combination of agenda.jobs() and job.disable()

enable(mongodb-native query)

Enables any jobs matching the passed mongodb-native query, allowing any matching jobs to be run by the Job Processor.

const numEnabled = await agenda.enable({ name: "pollExternalService" });

Similar to agenda.cancel(), this functionality can be acheived with a combination of agenda.jobs() and job.enable()

purge()

Removes all jobs in the database without defined behaviors. Useful if you change a definition name and want to remove old jobs. Returns a Promise resolving to the number of removed jobs, or rejecting on error.

IMPORTANT: Do not run this before you finish defining all of your jobs. If you do, you will nuke your database of jobs.

const numRemoved = await agenda.purge();

Starting the job processor

To get agenda to start processing jobs from the database you must start it. This will schedule an interval (based on processEvery) to check for new jobs and run them. You can also stop the queue.

start

Starts the job queue processing, checking processEvery time to see if there are new jobs. Must be called after processEvery, and before any job scheduling (e.g. every).

stop

Stops the job queue processing. Unlocks currently running jobs.

This can be very useful for graceful shutdowns so that currently running/grabbed jobs are abandoned so that other job queues can grab them / they are unlocked should the job queue start again. Here is an example of how to do a graceful shutdown.

async function graceful() {
  await agenda.stop();
  process.exit(0);
}

process.on("SIGTERM", graceful);
process.on("SIGINT", graceful);

close(force)

Closes database connection. You don't normally have to do this, but it might be useful for testing purposes.

Using force boolean you can force close connection.

Read more from Node.js MongoDB Driver API

await agenda.close({ force: true });

Multiple job processors

Sometimes you may want to have multiple node instances / machines process from the same queue. Agenda supports a locking mechanism to ensure that multiple queues don't process the same job.

You can configure the locking mechanism by specifying lockLifetime as an interval when defining the job.

agenda.define("someJob", { lockLifetime: 10000 }, (job, cb) => {
  // Do something in 10 seconds or less...
});

This will ensure that no other job processor (this one included) attempts to run the job again for the next 10 seconds. If you have a particularly long running job, you will want to specify a longer lockLifetime.

By default it is 10 minutes. Typically you shouldn't have a job that runs for 10 minutes, so this is really insurance should the job queue crash before the job is unlocked.

When a job is finished (i.e. the returned promise resolves/rejects or done is specified in the signature and done() is called), it will automatically unlock.

Manually working with a job

A job instance has many instance methods. All mutating methods must be followed with a call to await job.save() in order to persist the changes to the database.

repeatEvery(interval, [options])

Specifies an interval on which the job should repeat. The job runs at the time of defining as well in configured intervals, that is "run now and in intervals".

interval can be a human-readable format String, a cron format String, or a Number.

options is an optional argument containing:

options.timezone: should be a string as accepted by moment-timezone and is considered when using an interval in the cron string format.

options.skipImmediate: true | false (default) Setting this true will skip the immediate run. The first run will occur only in configured interval.

options.startDate: Date the first time the job runs, should be equal or after the start date.

options.endDate: Date the job should not repeat after the endDate. The job can run on the end-date itself, but not after that.

options.skipDays: humand readable string ('2 days'). After each run, it will skip the duration of 'skipDays'

job.repeatEvery("10 minutes");
await job.save();
job.repeatEvery("3 minutes", {
  skipImmediate: true,
});
await job.save();
job.repeatEvery("0 6 * * *", {
  timezone: "America/New_York",
});
await job.save();

repeatAt(time)

Specifies a time when the job should repeat. Possible values

job.repeatAt("3:30pm");
await job.save();

schedule(time)

Specifies the next time at which the job should run.

job.schedule("tomorrow at 6pm");
await job.save();

priority(priority)

Specifies the priority weighting of the job. Can be a number or a string from the above priority table.

job.priority("low");
await job.save();

setShouldSaveResult(setShouldSaveResult)

Specifies whether the result of the job should also be stored in the database. Defaults to false.

job.setShouldSaveResult(true);
await job.save();

The data returned by the job will be available on the result attribute after it succeeded and got retrieved again from the database, e.g. via agenda.jobs(...) or through the success job event).

unique(properties, [options])

Ensure that only one instance of this job exists with the specified properties

options is an optional argument which can overwrite the defaults. It can take the following:

  • insertOnly: boolean will prevent any properties from persisting if the job already exists. Defaults to false.
job.unique({ "data.type": "active", "data.userId": "123", nextRunAt: date });
await job.save();

IMPORTANT: To guarantee uniqueness as well as avoid high CPU usage by MongoDB make sure to create a unique index on the used fields, like name, data.type and data.userId for the example above.

fail(reason)

Sets job.attrs.failedAt to now, and sets job.attrs.failReason to reason.

Optionally, reason can be an error, in which case job.attrs.failReason will be set to error.message

job.fail("insufficient disk space");
// or
job.fail(new Error("insufficient disk space"));
await job.save();

run(callback)

Runs the given job and calls callback(err, job) upon completion. Normally you never need to call this manually.

job.run((err, job) => {
  console.log("I don't know why you would need to do this...");
});

save()

Saves the job.attrs into the database. Returns a Promise resolving to a Job instance, or rejecting on error.

try {
  await job.save();
  console.log("Successfully saved job to collection");
} catch (e) {
  console.error("Error saving job to collection");
}

remove()

Removes the job from the database. Returns a Promise resolving to the number of jobs removed, or rejecting on error.

try {
  await job.remove();
  console.log("Successfully removed job from collection");
} catch (e) {
  console.error("Error removing job from collection");
}

disable()

Disables the job. Upcoming runs won't execute.

enable()

Enables the job if it got disabled before. Upcoming runs will execute.

touch()

Resets the lock on the job. Useful to indicate that the job hasn't timed out when you have very long running jobs. The call returns a promise that resolves when the job's lock has been renewed.

agenda.define("super long job", async (job) => {
  await doSomeLongTask();
  await job.touch();
  await doAnotherLongTask();
  await job.touch();
  await finishOurLongTasks();
});

Job Queue Events

An instance of an agenda will emit the following events:

  • start - called just before a job starts
  • start:job name - called just before the specified job starts
agenda.on("start", (job) => {
  console.log("Job %s starting", job.attrs.name);
});
  • complete - called when a job finishes, regardless of if it succeeds or fails
  • complete:job name - called when a job finishes, regardless of if it succeeds or fails
agenda.on("complete", (job) => {
  console.log(`Job ${job.attrs.name} finished`);
});
  • success - called when a job finishes successfully
  • success:job name - called when a job finishes successfully
agenda.on("success:send email", (job) => {
  console.log(`Sent Email Successfully to ${job.attrs.data.to}`);
});
  • fail - called when a job throws an error
  • fail:job name - called when a job throws an error
agenda.on("fail:send email", (err, job) => {
  console.log(`Job failed with error: ${err.message}`);
});

Frequently Asked Questions

What is the order in which jobs run?

Jobs are run with priority in a first in first out order (so they will be run in the order they were scheduled AND with respect to highest priority).

For example, if we have two jobs named "send-email" queued (both with the same priority), and the first job is queued at 3:00 PM and second job is queued at 3:05 PM with the same priority value, then the first job will run first if we start to send "send-email" jobs at 3:10 PM. However if the first job has a priority of 5 and the second job has a priority of 10, then the second will run first (priority takes precedence) at 3:10 PM.

The default MongoDB sort object is { nextRunAt: 1, priority: -1 } and can be changed through the option sort when configuring Agenda.

What is the difference between lockLimit and maxConcurrency?

Agenda will lock jobs 1 by one, setting the lockedAt property in mongoDB, and creating an instance of the Job class which it caches into the _lockedJobs array. This defaults to having no limit, but can be managed using lockLimit. If all jobs will need to be run before agenda's next interval (set via agenda.processEvery), then agenda will attempt to lock all jobs.

Agenda will also pull jobs from _lockedJobs and into _runningJobs. These jobs are actively being worked on by user code, and this is limited by maxConcurrency (defaults to 20).

If you have multiple instances of agenda processing the same job definition with a fast repeat time you may find they get unevenly loaded. This is because they will compete to lock as many jobs as possible, even if they don't have enough concurrency to process them. This can be resolved by tweaking the maxConcurrency and lockLimit properties.

Sample Project Structure?

Agenda doesn't have a preferred project structure and leaves it to the user to choose how they would like to use it. That being said, you can check out the example project structure below.

Can I Donate?

Thanks! I'm flattered, but it's really not necessary. If you really want to, you can find my gittip here.

Web Interface?

Agenda itself does not have a web interface built in but we do offer stand-alone web interface Agendash:

Agendash interface

Mongo vs Redis

The decision to use Mongo instead of Redis is intentional. Redis is often used for non-essential data (such as sessions) and without configuration doesn't guarantee the same level of persistence as Mongo (should the server need to be restarted/crash).

Agenda decides to focus on persistence without requiring special configuration of Redis (thereby degrading the performance of the Redis server on non-critical data, such as sessions).

Ultimately if enough people want a Redis driver instead of Mongo, I will write one. (Please open an issue requesting it). For now, Agenda decided to focus on guaranteed persistence.

Spawning / forking processes

Ultimately Agenda can work from a single job queue across multiple machines, node processes, or forks. If you are interested in having more than one worker, Bars3s has written up a fantastic example of how one might do it:

const cluster = require("cluster");
const os = require("os");

const httpServer = require("./app/http-server");
const jobWorker = require("./app/job-worker");

const jobWorkers = [];
const webWorkers = [];

if (cluster.isMaster) {
  const cpuCount = os.cpus().length;
  // Create a worker for each CPU
  for (let i = 0; i < cpuCount; i += 1) {
    addJobWorker();
    addWebWorker();
  }

  cluster.on("exit", (worker, code, signal) => {
    if (jobWorkers.indexOf(worker.id) !== -1) {
      console.log(
        `job worker ${worker.process.pid} exited (signal: ${signal}). Trying to respawn...`
      );
      removeJobWorker(worker.id);
      addJobWorker();
    }

    if (webWorkers.indexOf(worker.id) !== -1) {
      console.log(
        `http worker ${worker.process.pid} exited (signal: ${signal}). Trying to respawn...`
      );
      removeWebWorker(worker.id);
      addWebWorker();
    }
  });
} else {
  if (process.env.web) {
    console.log(`start http server: ${cluster.worker.id}`);
    // Initialize the http server here
    httpServer.start();
  }

  if (process.env.job) {
    console.log(`start job server: ${cluster.worker.id}`);
    // Initialize the Agenda here
    jobWorker.start();
  }
}

function addWebWorker() {
  webWorkers.push(cluster.fork({ web: 1 }).id);
}

function addJobWorker() {
  jobWorkers.push(cluster.fork({ job: 1 }).id);
}

function removeWebWorker(id) {
  webWorkers.splice(webWorkers.indexOf(id), 1);
}

function removeJobWorker(id) {
  jobWorkers.splice(jobWorkers.indexOf(id), 1);
}

Recovering lost Mongo connections ("auto_reconnect")

Agenda is configured by default to automatically reconnect indefinitely, emitting an error event when no connection is available on each process tick, allowing you to restore the Mongo instance without having to restart the application.

However, if you are using an existing Mongo client you'll need to configure the reconnectTries and reconnectInterval connection settings manually, otherwise you'll find that Agenda will throw an error with the message "MongoDB connection is not recoverable, application restart required" if the connection cannot be recovered within 30 seconds.

Example Project Structure

Agenda will only process jobs that it has definitions for. This allows you to selectively choose which jobs a given agenda will process.

Consider the following project structure, which allows us to share models with the rest of our code base, and specify which jobs a worker processes, if any at all.

- server.js
- worker.js
lib/
  - agenda.js
  controllers/
    - user-controller.js
  jobs/
    - email.js
    - video-processing.js
    - image-processing.js
   models/
     - user-model.js
     - blog-post.model.js

Sample job processor (eg. jobs/email.js)

let email = require("some-email-lib"),
  User = require("../models/user-model.js");

module.exports = function (agenda) {
  agenda.define("registration email", async (job) => {
    const user = await User.get(job.attrs.data.userId);
    await email(
      user.email(),
      "Thanks for registering",
      "Thanks for registering " + user.name()
    );
  });

  agenda.define("reset password", async (job) => {
    // Etc
  });

  // More email related jobs
};

lib/agenda.js

const Agenda = require("agenda");

const connectionOpts = {
  db: { address: "localhost:27017/agenda-test", collection: "agendaJobs" },
};

const agenda = new Agenda(connectionOpts);

const jobTypes = process.env.JOB_TYPES ? process.env.JOB_TYPES.split(",") : [];

jobTypes.forEach((type) => {
  require("./jobs/" + type)(agenda);
});

if (jobTypes.length) {
  agenda.start(); // Returns a promise, which should be handled appropriately
}

module.exports = agenda;

lib/controllers/user-controller.js

let app = express(),
  User = require("../models/user-model"),
  agenda = require("../worker.js");

app.post("/users", (req, res, next) => {
  const user = new User(req.body);
  user.save((err) => {
    if (err) {
      return next(err);
    }
    agenda.now("registration email", { userId: user.primary() });
    res.send(201, user.toJson());
  });
});

worker.js

require("./lib/agenda.js");

Now you can do the following in your project:

node server.js

Fire up an instance with no JOB_TYPES, giving you the ability to process jobs, but not wasting resources processing jobs.

JOB_TYPES=email node server.js

Allow your http server to process email jobs.

JOB_TYPES=email node worker.js

Fire up an instance that processes email jobs.

JOB_TYPES=video-processing,image-processing node worker.js

Fire up an instance that processes video-processing/image-processing jobs. Good for a heavy hitting server.

Debugging Issues

If you think you have encountered a bug, please feel free to report it here:

Submit Issue

Please provide us with as much details as possible such as:

  • Agenda version
  • Environment (OSX, Linux, Windows, etc)
  • Small description of what happened
  • Any relevant stack track
  • Agenda logs (see below)

To turn on logging, please set your DEBUG env variable like so:

  • OSX: DEBUG="agenda:*" ts-node src/index.js
  • Linux: DEBUG="agenda:*" ts-node src/index.js
  • Windows CMD: set DEBUG=agenda:*
  • Windows PowerShell: $env:DEBUG = "agenda:*"

While not necessary, attaching a text file with this debug information would be extremely useful in debugging certain issues and is encouraged.

Known Issues

"Multiple order-by items are not supported. Please specify a single order-by item."

When running Agenda on Azure cosmosDB, you might run into this issue caused by Agenda's sort query used for finding and locking the next job. To fix this, you can pass custom sort option: sort: { nextRunAt: 1 }

Acknowledgements

Author: Agenda
Source Code: https://github.com/agenda/agenda 
License: View license

#node #task #cron 

What is GEEK

Buddha Community

Agenda: MongoDB-backed Job Scheduling

Rufus Scheduler: Job Scheduler for Ruby (at, Cron, in and Every Jobs)

rufus-scheduler

Job scheduler for Ruby (at, cron, in and every jobs).

It uses threads.

Note: maybe are you looking for the README of rufus-scheduler 2.x? (especially if you're using Dashing which is stuck on rufus-scheduler 2.0.24)

Quickstart:

# quickstart.rb

require 'rufus-scheduler'

scheduler = Rufus::Scheduler.new

scheduler.in '3s' do
  puts 'Hello... Rufus'
end

scheduler.join
  #
  # let the current thread join the scheduler thread
  #
  # (please note that this join should be removed when scheduling
  # in a web application (Rails and friends) initializer)

(run with ruby quickstart.rb)

Various forms of scheduling are supported:

require 'rufus-scheduler'

scheduler = Rufus::Scheduler.new

# ...

scheduler.in '10d' do
  # do something in 10 days
end

scheduler.at '2030/12/12 23:30:00' do
  # do something at a given point in time
end

scheduler.every '3h' do
  # do something every 3 hours
end
scheduler.every '3h10m' do
  # do something every 3 hours and 10 minutes
end

scheduler.cron '5 0 * * *' do
  # do something every day, five minutes after midnight
  # (see "man 5 crontab" in your terminal)
end

# ...

Rufus-scheduler uses fugit for parsing time strings, et-orbi for pairing time and tzinfo timezones.

non-features

Rufus-scheduler (out of the box) is an in-process, in-memory scheduler. It uses threads.

It does not persist your schedules. When the process is gone and the scheduler instance with it, the schedules are gone.

A rufus-scheduler instance will go on scheduling while it is present among the objects in a Ruby process. To make it stop scheduling you have to call its #shutdown method.

related and similar gems

  • Whenever - let cron call back your Ruby code, trusted and reliable cron drives your schedule
  • ruby-clock - a clock process / job scheduler for Ruby
  • Clockwork - rufus-scheduler inspired gem
  • Crono - an in-Rails cron scheduler
  • PerfectSched - highly available distributed cron built on Sequel and more

(please note: rufus-scheduler is not a cron replacement)

note about the 3.0 line

It's a complete rewrite of rufus-scheduler.

There is no EventMachine-based scheduler anymore.

I don't know what this Ruby thing is, where are my Rails?

I'll drive you right to the tracks.

notable changes:

  • As said, no more EventMachine-based scheduler
  • scheduler.every('100') { will schedule every 100 seconds (previously, it would have been 0.1s). This aligns rufus-scheduler with Ruby's sleep(100)
  • The scheduler isn't catching the whole of Exception anymore, only StandardError
  • The error_handler is #on_error (instead of #on_exception), by default it now prints the details of the error to $stderr (used to be $stdout)
  • Rufus::Scheduler::TimeOutError renamed to Rufus::Scheduler::TimeoutError
  • Introduction of "interval" jobs. Whereas "every" jobs are like "every 10 minutes, do this", interval jobs are like "do that, then wait for 10 minutes, then do that again, and so on"
  • Introduction of a lockfile: true/filename mechanism to prevent multiple schedulers from executing
  • "discard_past" is on by default. If the scheduler (its host) sleeps for 1 hour and a every '10m' job is on, it will trigger once at wakeup, not 6 times (discard_past was false by default in rufus-scheduler 2.x). No intention to re-introduce discard_past: false in 3.0 for now.
  • Introduction of Scheduler #on_pre_trigger and #on_post_trigger callback points

getting help

So you need help. People can help you, but first help them help you, and don't waste their time. Provide a complete description of the issue. If it works on A but not on B and others have to ask you: "so what is different between A and B" you are wasting everyone's time.

"hello", "please" and "thanks" are not swear words.

Go read how to report bugs effectively, twice.

Update: help_help.md might help help you.

on Gitter

You can find help via chat over at https://gitter.im/floraison/fugit. It's fugit, et-orbi, and rufus-scheduler combined chat room.

Please be courteous.

issues

Yes, issues can be reported in rufus-scheduler issues, I'd actually prefer bugs in there. If there is nothing wrong with rufus-scheduler, a Stack Overflow question is better.

faq

scheduling

Rufus-scheduler supports five kinds of jobs. in, at, every, interval and cron jobs.

Most of the rufus-scheduler examples show block scheduling, but it's also OK to schedule handler instances or handler classes.

in, at, every, interval, cron

In and at jobs trigger once.

require 'rufus-scheduler'

scheduler = Rufus::Scheduler.new

scheduler.in '10d' do
  puts "10 days reminder for review X!"
end

scheduler.at '2014/12/24 2000' do
  puts "merry xmas!"
end

In jobs are scheduled with a time interval, they trigger after that time elapsed. At jobs are scheduled with a point in time, they trigger when that point in time is reached (better to choose a point in the future).

Every, interval and cron jobs trigger repeatedly.

require 'rufus-scheduler'

scheduler = Rufus::Scheduler.new

scheduler.every '3h' do
  puts "change the oil filter!"
end

scheduler.interval '2h' do
  puts "thinking..."
  puts sleep(rand * 1000)
  puts "thought."
end

scheduler.cron '00 09 * * *' do
  puts "it's 9am! good morning!"
end

Every jobs try hard to trigger following the frequency they were scheduled with.

Interval jobs trigger, execute and then trigger again after the interval elapsed. (every jobs time between trigger times, interval jobs time between trigger termination and the next trigger start).

Cron jobs are based on the venerable cron utility (man 5 crontab). They trigger following a pattern given in (almost) the same language cron uses.

 

#schedule_x vs #x

schedule_in, schedule_at, schedule_cron, etc will return the new Job instance.

in, at, cron will return the new Job instance's id (a String).

job_id =
  scheduler.in '10d' do
    # ...
  end
job = scheduler.job(job_id)

# versus

job =
  scheduler.schedule_in '10d' do
    # ...
  end

# also

job =
  scheduler.in '10d', job: true do
    # ...
  end

#schedule and #repeat

Sometimes it pays to be less verbose.

The #schedule methods schedules an at, in or cron job. It just decides based on its input. It returns the Job instance.

scheduler.schedule '10d' do; end.class
  # => Rufus::Scheduler::InJob

scheduler.schedule '2013/12/12 12:30' do; end.class
  # => Rufus::Scheduler::AtJob

scheduler.schedule '* * * * *' do; end.class
  # => Rufus::Scheduler::CronJob

The #repeat method schedules and returns an EveryJob or a CronJob.

scheduler.repeat '10d' do; end.class
  # => Rufus::Scheduler::EveryJob

scheduler.repeat '* * * * *' do; end.class
  # => Rufus::Scheduler::CronJob

(Yes, no combination here gives back an IntervalJob).

schedule blocks arguments (job, time)

A schedule block may be given 0, 1 or 2 arguments.

The first argument is "job", it's simply the Job instance involved. It might be useful if the job is to be unscheduled for some reason.

scheduler.every '10m' do |job|

  status = determine_pie_status

  if status == 'burnt' || status == 'cooked'
    stop_oven
    takeout_pie
    job.unschedule
  end
end

The second argument is "time", it's the time when the job got cleared for triggering (not Time.now).

Note that time is the time when the job got cleared for triggering. If there are mutexes involved, now = mutex_wait_time + time...

"every" jobs and changing the next_time in-flight

It's OK to change the next_time of an every job in-flight:

scheduler.every '10m' do |job|

  # ...

  status = determine_pie_status

  job.next_time = Time.now + 30 * 60 if status == 'burnt'
    #
    # if burnt, wait 30 minutes for the oven to cool a bit
end

It should work as well with cron jobs, not so with interval jobs whose next_time is computed after their block ends its current run.

scheduling handler instances

It's OK to pass any object, as long as it responds to #call(), when scheduling:

class Handler
  def self.call(job, time)
    p "- Handler called for #{job.id} at #{time}"
  end
end

scheduler.in '10d', Handler

# or

class OtherHandler
  def initialize(name)
    @name = name
  end
  def call(job, time)
    p "* #{time} - Handler #{name.inspect} called for #{job.id}"
  end
end

oh = OtherHandler.new('Doe')

scheduler.every '10m', oh
scheduler.in '3d5m', oh

The call method must accept 2 (job, time), 1 (job) or 0 arguments.

Note that time is the time when the job got cleared for triggering. If there are mutexes involved, now = mutex_wait_time + time...

scheduling handler classes

One can pass a handler class to rufus-scheduler when scheduling. Rufus will instantiate it and that instance will be available via job#handler.

class MyHandler
  attr_reader :count
  def initialize
    @count = 0
  end
  def call(job)
    @count += 1
    puts ". #{self.class} called at #{Time.now} (#{@count})"
  end
end

job = scheduler.schedule_every '35m', MyHandler

job.handler
  # => #<MyHandler:0x000000021034f0>
job.handler.count
  # => 0

If you want to keep that "block feeling":

job_id =
  scheduler.every '10m', Class.new do
    def call(job)
      puts ". hello #{self.inspect} at #{Time.now}"
    end
  end

pause and resume the scheduler

The scheduler can be paused via the #pause and #resume methods. One can determine if the scheduler is currently paused by calling #paused?.

While paused, the scheduler still accepts schedules, but no schedule will get triggered as long as #resume isn't called.

job options

name: string

Sets the name of the job.

scheduler.cron '*/15 8 * * *', name: 'Robert' do |job|
  puts "A, it's #{Time.now} and my name is #{job.name}"
end

job1 =
  scheduler.schedule_cron '*/30 9 * * *', n: 'temporary' do |job|
    puts "B, it's #{Time.now} and my name is #{job.name}"
  end
# ...
job1.name = 'Beowulf'

blocking: true

By default, jobs are triggered in their own, new threads. When blocking: true, the job is triggered in the scheduler thread (a new thread is not created). Yes, while a blocking job is running, the scheduler is not scheduling.

overlap: false

Since, by default, jobs are triggered in their own new threads, job instances might overlap. For example, a job that takes 10 minutes and is scheduled every 7 minutes will have overlaps.

To prevent overlap, one can set overlap: false. Such a job will not trigger if one of its instances is already running.

The :overlap option is considered before the :mutex option when the scheduler is reviewing jobs for triggering.

mutex: mutex_instance / mutex_name / array of mutexes

When a job with a mutex triggers, the job's block is executed with the mutex around it, preventing other jobs with the same mutex from entering (it makes the other jobs wait until it exits the mutex).

This is different from overlap: false, which is, first, limited to instances of the same job, and, second, doesn't make the incoming job instance block/wait but give up.

:mutex accepts a mutex instance or a mutex name (String). It also accept an array of mutex names / mutex instances. It allows for complex relations between jobs.

Array of mutexes: original idea and implementation by Rainux Luo

Note: creating lots of different mutexes is OK. Rufus-scheduler will place them in its Scheduler#mutexes hash... And they won't get garbage collected.

The :overlap option is considered before the :mutex option when the scheduler is reviewing jobs for triggering.

timeout: duration or point in time

It's OK to specify a timeout when scheduling some work. After the time specified, it gets interrupted via a Rufus::Scheduler::TimeoutError.

scheduler.in '10d', timeout: '1d' do
  begin
    # ... do something
  rescue Rufus::Scheduler::TimeoutError
    # ... that something got interrupted after 1 day
  end
end

The :timeout option accepts either a duration (like "1d" or "2w3d") or a point in time (like "2013/12/12 12:00").

:first_at, :first_in, :first, :first_time

This option is for repeat jobs (cron / every) only.

It's used to specify the first time after which the repeat job should trigger for the first time.

In the case of an "every" job, this will be the first time (modulo the scheduler frequency) the job triggers. For a "cron" job as well, the :first will point to the first time the job has to trigger, the following trigger times are then determined by the cron string.

scheduler.every '2d', first_at: Time.now + 10 * 3600 do
  # ... every two days, but start in 10 hours
end

scheduler.every '2d', first_in: '10h' do
  # ... every two days, but start in 10 hours
end

scheduler.cron '00 14 * * *', first_in: '3d' do
  # ... every day at 14h00, but start after 3 * 24 hours
end

:first, :first_at and :first_in all accept a point in time or a duration (number or time string). Use the symbol you think makes your schedule more readable.

Note: it's OK to change the first_at (a Time instance) directly:

job.first_at = Time.now + 10
job.first_at = Rufus::Scheduler.parse('2029-12-12')

The first argument (in all its flavours) accepts a :now or :immediately value. That schedules the first occurrence for immediate triggering. Consider:

require 'rufus-scheduler'

s = Rufus::Scheduler.new

n = Time.now; p [ :scheduled_at, n, n.to_f ]

s.every '3s', first: :now do
  n = Time.now; p [ :in, n, n.to_f ]
end

s.join

that'll output something like:

[:scheduled_at, 2014-01-22 22:21:21 +0900, 1390396881.344438]
[:in, 2014-01-22 22:21:21 +0900, 1390396881.6453865]
[:in, 2014-01-22 22:21:24 +0900, 1390396884.648807]
[:in, 2014-01-22 22:21:27 +0900, 1390396887.651686]
[:in, 2014-01-22 22:21:30 +0900, 1390396890.6571937]
...

:last_at, :last_in, :last

This option is for repeat jobs (cron / every) only.

It indicates the point in time after which the job should unschedule itself.

scheduler.cron '5 23 * * *', last_in: '10d' do
  # ... do something every evening at 23:05 for 10 days
end

scheduler.every '10m', last_at: Time.now + 10 * 3600 do
  # ... do something every 10 minutes for 10 hours
end

scheduler.every '10m', last_in: 10 * 3600 do
  # ... do something every 10 minutes for 10 hours
end

:last, :last_at and :last_in all accept a point in time or a duration (number or time string). Use the symbol you think makes your schedule more readable.

Note: it's OK to change the last_at (nil or a Time instance) directly:

job.last_at = nil
  # remove the "last" bound

job.last_at = Rufus::Scheduler.parse('2029-12-12')
  # set the last bound

times: nb of times (before auto-unscheduling)

One can tell how many times a repeat job (CronJob or EveryJob) is to execute before unscheduling by itself.

scheduler.every '2d', times: 10 do
  # ... do something every two days, but not more than 10 times
end

scheduler.cron '0 23 * * *', times: 31 do
  # ... do something every day at 23:00 but do it no more than 31 times
end

It's OK to assign nil to :times to make sure the repeat job is not limited. It's useful when the :times is determined at scheduling time.

scheduler.cron '0 23 * * *', times: (nolimit ? nil : 10) do
  # ...
end

The value set by :times is accessible in the job. It can be modified anytime.

job =
  scheduler.cron '0 23 * * *' do
    # ...
  end

# later on...

job.times = 10
  # 10 days and it will be over

Job methods

When calling a schedule method, the id (String) of the job is returned. Longer schedule methods return Job instances directly. Calling the shorter schedule methods with the job: true also returns Job instances instead of Job ids (Strings).

  require 'rufus-scheduler'

  scheduler = Rufus::Scheduler.new

  job_id =
    scheduler.in '10d' do
      # ...
    end

  job =
    scheduler.schedule_in '1w' do
      # ...
    end

  job =
    scheduler.in '1w', job: true do
      # ...
    end

Those Job instances have a few interesting methods / properties:

id, job_id

Returns the job id.

job = scheduler.schedule_in('10d') do; end
job.id
  # => "in_1374072446.8923042_0.0_0"

scheduler

Returns the scheduler instance itself.

opts

Returns the options passed at the Job creation.

job = scheduler.schedule_in('10d', tag: 'hello') do; end
job.opts
  # => { :tag => 'hello' }

original

Returns the original schedule.

job = scheduler.schedule_in('10d', tag: 'hello') do; end
job.original
  # => '10d'

callable, handler

callable() returns the scheduled block (or the call method of the callable object passed in lieu of a block)

handler() returns nil if a block was scheduled and the instance scheduled otherwise.

# when passing a block

job =
  scheduler.schedule_in('10d') do
    # ...
  end

job.handler
  # => nil
job.callable
  # => #<Proc:0x00000001dc6f58@/home/jmettraux/whatever.rb:115>

and

# when passing something else than a block

class MyHandler
  attr_reader :counter
  def initialize
    @counter = 0
  end
  def call(job, time)
    @counter = @counter + 1
  end
end

job = scheduler.schedule_in('10d', MyHandler.new)

job.handler
  # => #<Method: MyHandler#call>
job.callable
  # => #<MyHandler:0x0000000163ae88 @counter=0>

source_location

Added to rufus-scheduler 3.8.0.

Returns the array [ 'path/to/file.rb', 123 ] like Proc#source_location does.

require 'rufus-scheduler'

scheduler = Rufus::Scheduler.new

job = scheduler.schedule_every('2h') { p Time.now }

p job.source_location
  # ==> [ '/home/jmettraux/rufus-scheduler/test.rb', 6 ]

scheduled_at

Returns the Time instance when the job got created.

job = scheduler.schedule_in('10d', tag: 'hello') do; end
job.scheduled_at
  # => 2013-07-17 23:48:54 +0900

last_time

Returns the last time the job triggered (is usually nil for AtJob and InJob).

job = scheduler.schedule_every('10s') do; end

job.scheduled_at
  # => 2013-07-17 23:48:54 +0900
job.last_time
  # => nil (since we've just scheduled it)

# after 10 seconds

job.scheduled_at
  # => 2013-07-17 23:48:54 +0900 (same as above)
job.last_time
  # => 2013-07-17 23:49:04 +0900

previous_time

Returns the previous #next_time

scheduler.every('10s') do |job|
  puts "job scheduled for #{job.previous_time} triggered at #{Time.now}"
  puts "next time will be around #{job.next_time}"
  puts "."
end

last_work_time, mean_work_time

The job keeps track of how long its work was in the last_work_time attribute. For a one time job (in, at) it's probably not very useful.

The attribute mean_work_time contains a computed mean work time. It's recomputed after every run (if it's a repeat job).

next_times(n)

Returns an array of EtOrbi::EoTime instances (Time instances with a designated time zone), listing the n next occurrences for this job.

Please note that for "interval" jobs, a mean work time is computed each time and it's used by this #next_times(n) method to approximate the next times beyond the immediate next time.

unschedule

Unschedule the job, preventing it from firing again and removing it from the schedule. This doesn't prevent a running thread for this job to run until its end.

threads

Returns the list of threads currently "hosting" runs of this Job instance.

kill

Interrupts all the work threads currently running for this job instance. They discard their work and are free for their next run (of whatever job).

Note: this doesn't unschedule the Job instance.

Note: if the job is pooled for another run, a free work thread will probably pick up that next run and the job will appear as running again. You'd have to unschedule and kill to make sure the job doesn't run again.

running?

Returns true if there is at least one running Thread hosting a run of this Job instance.

scheduled?

Returns true if the job is scheduled (is due to trigger). For repeat jobs it should return true until the job gets unscheduled. "at" and "in" jobs will respond with false as soon as they start running (execution triggered).

pause, resume, paused?, paused_at

These four methods are only available to CronJob, EveryJob and IntervalJob instances. One can pause or resume such jobs thanks to these methods.

job =
  scheduler.schedule_every('10s') do
    # ...
  end

job.pause
  # => 2013-07-20 01:22:22 +0900
job.paused?
  # => true
job.paused_at
  # => 2013-07-20 01:22:22 +0900

job.resume
  # => nil

tags

Returns the list of tags attached to this Job instance.

By default, returns an empty array.

job = scheduler.schedule_in('10d') do; end
job.tags
  # => []

job = scheduler.schedule_in('10d', tag: 'hello') do; end
job.tags
  # => [ 'hello' ]

[]=, [], key?, has_key?, keys, values, and entries

Threads have thread-local variables, similarly Rufus-scheduler jobs have job-local variables. Those are more like a dict with thread-safe access.

job =
  @scheduler.schedule_every '1s' do |job|
    job[:timestamp] = Time.now.to_f
    job[:counter] ||= 0
    job[:counter] += 1
  end

sleep 3.6

job[:counter]
  # => 3

job.key?(:timestamp) # => true
job.has_key?(:timestamp) # => true
job.keys # => [ :timestamp, :counter ]

Locals can be set at schedule time:

job0 =
  @scheduler.schedule_cron '*/15 12 * * *', locals: { a: 0 } do
    # ...
  end
job1 =
  @scheduler.schedule_cron '*/15 13 * * *', l: { a: 1 } do
    # ...
  end

One can fetch the Hash directly with Job#locals. Of course, direct manipulation is not thread-safe.

job.locals.entries do |k, v|
  p "#{k}: #{v}"
end

call

Job instances have a #call method. It simply calls the scheduled block or callable immediately.

job =
  @scheduler.schedule_every '10m' do |job|
    # ...
  end

job.call

Warning: the Scheduler#on_error handler is not involved. Error handling is the responsibility of the caller.

If the call has to be rescued by the error handler of the scheduler, call(true) might help:

require 'rufus-scheduler'

s = Rufus::Scheduler.new

def s.on_error(job, err)
  if job
    p [ 'error in scheduled job', job.class, job.original, err.message ]
  else
    p [ 'error while scheduling', err.message ]
  end
rescue
  p $!
end

job =
  s.schedule_in('1d') do
    fail 'again'
  end

job.call(true)
  #
  # true lets the error_handler deal with error in the job call

AtJob and InJob methods

time

Returns when the job will trigger (hopefully).

next_time

An alias for time.

EveryJob, IntervalJob and CronJob methods

next_time

Returns the next time the job will trigger (hopefully).

count

Returns how many times the job fired.

EveryJob methods

frequency

It returns the scheduling frequency. For a job scheduled "every 20s", it's 20.

It's used to determine if the job frequency is higher than the scheduler frequency (it raises an ArgumentError if that is the case).

IntervalJob methods

interval

Returns the interval scheduled between each execution of the job.

Every jobs use a time duration between each start of their execution, while interval jobs use a time duration between the end of an execution and the start of the next.

CronJob methods

brute_frequency

An expensive method to run, it's brute. It caches its results. By default it runs for 2017 (a non leap-year).

  require 'rufus-scheduler'

  Rufus::Scheduler.parse('* * * * *').brute_frequency
    #
    # => #<Fugit::Cron::Frequency:0x00007fdf4520c5e8
    #      @span=31536000.0, @delta_min=60, @delta_max=60,
    #      @occurrences=525600, @span_years=1.0, @yearly_occurrences=525600.0>
      #
      # Occurs 525600 times in a span of 1 year (2017) and 1 day.
      # There are least 60 seconds between "triggers" and at most 60 seconds.

  Rufus::Scheduler.parse('0 12 * * *').brute_frequency
    # => #<Fugit::Cron::Frequency:0x00007fdf451ec6d0
    #      @span=31536000.0, @delta_min=86400, @delta_max=86400,
    #      @occurrences=365, @span_years=1.0, @yearly_occurrences=365.0>
  Rufus::Scheduler.parse('0 12 * * *').brute_frequency.to_debug_s
    # => "dmin: 1D, dmax: 1D, ocs: 365, spn: 52W1D, spnys: 1, yocs: 365"
      #
      # 365 occurrences, at most 1 day between each, at least 1 day.

The CronJob#frequency method found in rufus-scheduler < 3.5 has been retired.

looking up jobs

Scheduler#job(job_id)

The scheduler #job(job_id) method can be used to look up Job instances.

  require 'rufus-scheduler'

  scheduler = Rufus::Scheduler.new

  job_id =
    scheduler.in '10d' do
      # ...
    end

  # later on...

  job = scheduler.job(job_id)

Scheduler #jobs #at_jobs #in_jobs #every_jobs #interval_jobs and #cron_jobs

Are methods for looking up lists of scheduled Job instances.

Here is an example:

  #
  # let's unschedule all the at jobs

  scheduler.at_jobs.each(&:unschedule)

Scheduler#jobs(tag: / tags: x)

When scheduling a job, one can specify one or more tags attached to the job. These can be used to look up the job later on.

  scheduler.in '10d', tag: 'main_process' do
    # ...
  end
  scheduler.in '10d', tags: [ 'main_process', 'side_dish' ] do
    # ...
  end

  # ...

  jobs = scheduler.jobs(tag: 'main_process')
    # find all the jobs with the 'main_process' tag

  jobs = scheduler.jobs(tags: [ 'main_process', 'side_dish' ]
    # find all the jobs with the 'main_process' AND 'side_dish' tags

Scheduler#running_jobs

Returns the list of Job instance that have currently running instances.

Whereas other "_jobs" method scan the scheduled job list, this method scans the thread list to find the job. It thus comprises jobs that are running but are not scheduled anymore (that happens for at and in jobs).

misc Scheduler methods

Scheduler#unschedule(job_or_job_id)

Unschedule a job given directly or by its id.

Scheduler#shutdown

Shuts down the scheduler, ceases any scheduler/triggering activity.

Scheduler#shutdown(:wait)

Shuts down the scheduler, waits (blocks) until all the jobs cease running.

Scheduler#shutdown(wait: n)

Shuts down the scheduler, waits (blocks) at most n seconds until all the jobs cease running. (Jobs are killed after n seconds have elapsed).

Scheduler#shutdown(:kill)

Kills all the job (threads) and then shuts the scheduler down. Radical.

Scheduler#down?

Returns true if the scheduler has been shut down.

Scheduler#started_at

Returns the Time instance at which the scheduler got started.

Scheduler #uptime / #uptime_s

Returns since the count of seconds for which the scheduler has been running.

#uptime_s returns this count in a String easier to grasp for humans, like "3d12m45s123".

Scheduler#join

Lets the current thread join the scheduling thread in rufus-scheduler. The thread comes back when the scheduler gets shut down.

#join is mostly used in standalone scheduling script (or tiny one file examples). Calling #join from a web application initializer will probably hijack the main thread and prevent the web application from being served. Do not put a #join in such a web application initializer file.

Scheduler#threads

Returns all the threads associated with the scheduler, including the scheduler thread itself.

Scheduler#work_threads(query=:all/:active/:vacant)

Lists the work threads associated with the scheduler. The query option defaults to :all.

  • :all : all the work threads
  • :active : all the work threads currently running a Job
  • :vacant : all the work threads currently not running a Job

Note that the main schedule thread will be returned if it is currently running a Job (ie one of those blocking: true jobs).

Scheduler#scheduled?(job_or_job_id)

Returns true if the arg is a currently scheduled job (see Job#scheduled?).

Scheduler#occurrences(time0, time1)

Returns a hash { job => [ t0, t1, ... ] } mapping jobs to their potential trigger time within the [ time0, time1 ] span.

Please note that, for interval jobs, the #mean_work_time is used, so the result is only a prediction.

Scheduler#timeline(time0, time1)

Like #occurrences but returns a list [ [ t0, job0 ], [ t1, job1 ], ... ] of time + job pairs.

dealing with job errors

The easy, job-granular way of dealing with errors is to rescue and deal with them immediately. The two next sections show examples. Skip them for explanations on how to deal with errors at the scheduler level.

block jobs

As said, jobs could take care of their errors themselves.

scheduler.every '10m' do
  begin
    # do something that might fail...
  rescue => e
    $stderr.puts '-' * 80
    $stderr.puts e.message
    $stderr.puts e.stacktrace
    $stderr.puts '-' * 80
  end
end

callable jobs

Jobs are not only shrunk to blocks, here is how the above would look like with a dedicated class.

scheduler.every '10m', Class.new do
  def call(job)
    # do something that might fail...
  rescue => e
    $stderr.puts '-' * 80
    $stderr.puts e.message
    $stderr.puts e.stacktrace
    $stderr.puts '-' * 80
  end
end

TODO: talk about callable#on_error (if implemented)

(see scheduling handler instances and scheduling handler classes for more about those "callable jobs")

Rufus::Scheduler#stderr=

By default, rufus-scheduler intercepts all errors (that inherit from StandardError) and dumps abundant details to $stderr.

If, for example, you'd like to divert that flow to another file (descriptor), you can reassign $stderr for the current Ruby process

$stderr = File.open('/var/log/myapplication.log', 'ab')

or, you can limit that reassignement to the scheduler itself

scheduler.stderr = File.open('/var/log/myapplication.log', 'ab')

Rufus::Scheduler#on_error(job, error)

We've just seen that, by default, rufus-scheduler dumps error information to $stderr. If one needs to completely change what happens in case of error, it's OK to overwrite #on_error

def scheduler.on_error(job, error)

  Logger.warn("intercepted error in #{job.id}: #{error.message}")
end

On Rails, the on_error method redefinition might look like:

def scheduler.on_error(job, error)

  Rails.logger.error(
    "err#{error.object_id} rufus-scheduler intercepted #{error.inspect}" +
    " in job #{job.inspect}")
  error.backtrace.each_with_index do |line, i|
    Rails.logger.error(
      "err#{error.object_id} #{i}: #{line}")
  end
end

Callbacks

Rufus::Scheduler #on_pre_trigger and #on_post_trigger callbacks

One can bind callbacks before and after jobs trigger:

s = Rufus::Scheduler.new

def s.on_pre_trigger(job, trigger_time)
  puts "triggering job #{job.id}..."
end

def s.on_post_trigger(job, trigger_time)
  puts "triggered job #{job.id}."
end

s.every '1s' do
  # ...
end

The trigger_time is the time at which the job triggers. It might be a bit before Time.now.

Warning: these two callbacks are executed in the scheduler thread, not in the work threads (the threads where the job execution really happens).

Rufus::Scheduler#around_trigger

One can create an around callback which will wrap a job:

def s.around_trigger(job)
  t = Time.now
  puts "Starting job #{job.id}..."
  yield
  puts "job #{job.id} finished in #{Time.now-t} seconds."
end

The around callback is executed in the thread.

Rufus::Scheduler#on_pre_trigger as a guard

Returning false in on_pre_trigger will prevent the job from triggering. Returning anything else (nil, -1, true, ...) will let the job trigger.

Note: your business logic should go in the scheduled block itself (or the scheduled instance). Don't put business logic in on_pre_trigger. Return false for admin reasons (backend down, etc), not for business reasons that are tied to the job itself.

def s.on_pre_trigger(job, trigger_time)

  return false if Backend.down?

  puts "triggering job #{job.id}..."
end

Rufus::Scheduler.new options

:frequency

By default, rufus-scheduler sleeps 0.300 second between every step. At each step it checks for jobs to trigger and so on.

The :frequency option lets you change that 0.300 second to something else.

scheduler = Rufus::Scheduler.new(frequency: 5)

It's OK to use a time string to specify the frequency.

scheduler = Rufus::Scheduler.new(frequency: '2h10m')
  # this scheduler will sleep 2 hours and 10 minutes between every "step"

Use with care.

lockfile: "mylockfile.txt"

This feature only works on OSes that support the flock (man 2 flock) call.

Starting the scheduler with lockfile: '.rufus-scheduler.lock' will make the scheduler attempt to create and lock the file .rufus-scheduler.lock in the current working directory. If that fails, the scheduler will not start.

The idea is to guarantee only one scheduler (in a group of schedulers sharing the same lockfile) is running.

This is useful in environments where the Ruby process holding the scheduler gets started multiple times.

If the lockfile mechanism here is not sufficient, you can plug your custom mechanism. It's explained in advanced lock schemes below.

:scheduler_lock

(since rufus-scheduler 3.0.9)

The scheduler lock is an object that responds to #lock and #unlock. The scheduler calls #lock when starting up. If the answer is false, the scheduler stops its initialization work and won't schedule anything.

Here is a sample of a scheduler lock that only lets the scheduler on host "coffee.example.com" start:

class HostLock
  def initialize(lock_name)
    @lock_name = lock_name
  end
  def lock
    @lock_name == `hostname -f`.strip
  end
  def unlock
    true
  end
end

scheduler =
  Rufus::Scheduler.new(scheduler_lock: HostLock.new('coffee.example.com'))

By default, the scheduler_lock is an instance of Rufus::Scheduler::NullLock, with a #lock that returns true.

:trigger_lock

(since rufus-scheduler 3.0.9)

The trigger lock in an object that responds to #lock. The scheduler calls that method on the job lock right before triggering any job. If the answer is false, the trigger doesn't happen, the job is not done (at least not in this scheduler).

Here is a (stupid) PingLock example, it'll only trigger if an "other host" is not responding to ping. Do not use that in production, you don't want to fork a ping process for each trigger attempt...

class PingLock
  def initialize(other_host)
    @other_host = other_host
  end
  def lock
    ! system("ping -c 1 #{@other_host}")
  end
end

scheduler =
  Rufus::Scheduler.new(trigger_lock: PingLock.new('main.example.com'))

By default, the trigger_lock is an instance of Rufus::Scheduler::NullLock, with a #lock that always returns true.

As explained in advanced lock schemes, another way to tune that behaviour is by overriding the scheduler's #confirm_lock method. (You could also do that with an #on_pre_trigger callback).

:max_work_threads

In rufus-scheduler 2.x, by default, each job triggering received its own, brand new, thread of execution. In rufus-scheduler 3.x, execution happens in a pooled work thread. The max work thread count (the pool size) defaults to 28.

One can set this maximum value when starting the scheduler.

scheduler = Rufus::Scheduler.new(max_work_threads: 77)

It's OK to increase the :max_work_threads of a running scheduler.

scheduler.max_work_threads += 10

Rufus::Scheduler.singleton

Do not want to store a reference to your rufus-scheduler instance? Then Rufus::Scheduler.singleton can help, it returns a singleton instance of the scheduler, initialized the first time this class method is called.

Rufus::Scheduler.singleton.every '10s' { puts "hello, world!" }

It's OK to pass initialization arguments (like :frequency or :max_work_threads) but they will only be taken into account the first time .singleton is called.

Rufus::Scheduler.singleton(max_work_threads: 77)
Rufus::Scheduler.singleton(max_work_threads: 277) # no effect

The .s is a shortcut for .singleton.

Rufus::Scheduler.s.every '10s' { puts "hello, world!" }

advanced lock schemes

As seen above, rufus-scheduler proposes the :lockfile system out of the box. If in a group of schedulers only one is supposed to run, the lockfile mechanism prevents schedulers that have not set/created the lockfile from running.

There are situations where this is not sufficient.

By overriding #lock and #unlock, one can customize how schedulers lock.

This example was provided by Eric Lindvall:

class ZookeptScheduler < Rufus::Scheduler

  def initialize(zookeeper, opts={})
    @zk = zookeeper
    super(opts)
  end

  def lock
    @zk_locker = @zk.exclusive_locker('scheduler')
    @zk_locker.lock # returns true if the lock was acquired, false else
  end

  def unlock
    @zk_locker.unlock
  end

  def confirm_lock
    return false if down?
    @zk_locker.assert!
  rescue ZK::Exceptions::LockAssertionFailedError => e
    # we've lost the lock, shutdown (and return false to at least prevent
    # this job from triggering
    shutdown
    false
  end
end

This uses a zookeeper to make sure only one scheduler in a group of distributed schedulers runs.

The methods #lock and #unlock are overridden and #confirm_lock is provided, to make sure that the lock is still valid.

The #confirm_lock method is called right before a job triggers (if it is provided). The more generic callback #on_pre_trigger is called right after #confirm_lock.

:scheduler_lock and :trigger_lock

(introduced in rufus-scheduler 3.0.9).

Another way of prodiving #lock, #unlock and #confirm_lock to a rufus-scheduler is by using the :scheduler_lock and :trigger_lock options.

See :trigger_lock and :scheduler_lock.

The scheduler lock may be used to prevent a scheduler from starting, while a trigger lock prevents individual jobs from triggering (the scheduler goes on scheduling).

One has to be careful with what goes in #confirm_lock or in a trigger lock, as it gets called before each trigger.

Warning: you may think you're heading towards "high availability" by using a trigger lock and having lots of schedulers at hand. It may be so if you limit yourself to scheduling the same set of jobs at scheduler startup. But if you add schedules at runtime, they stay local to their scheduler. There is no magic that propagates the jobs to all the schedulers in your pack.

parsing cronlines and time strings

(Please note that fugit does the heavy-lifting parsing work for rufus-scheduler).

Rufus::Scheduler provides a class method .parse to parse time durations and cron strings. It's what it's using when receiving schedules. One can use it directly (no need to instantiate a Scheduler).

require 'rufus-scheduler'

Rufus::Scheduler.parse('1w2d')
  # => 777600.0
Rufus::Scheduler.parse('1.0w1.0d')
  # => 777600.0

Rufus::Scheduler.parse('Sun Nov 18 16:01:00 2012').strftime('%c')
  # => 'Sun Nov 18 16:01:00 2012'

Rufus::Scheduler.parse('Sun Nov 18 16:01:00 2012 Europe/Berlin').strftime('%c %z')
  # => 'Sun Nov 18 15:01:00 2012 +0000'

Rufus::Scheduler.parse(0.1)
  # => 0.1

Rufus::Scheduler.parse('* * * * *')
  # => #<Fugit::Cron:0x00007fb7a3045508
  #      @original="* * * * *", @cron_s=nil,
  #      @seconds=[0], @minutes=nil, @hours=nil, @monthdays=nil, @months=nil,
  #      @weekdays=nil, @zone=nil, @timezone=nil>

It returns a number when the input is a duration and a Fugit::Cron instance when the input is a cron string.

It will raise an ArgumentError if it can't parse the input.

Beyond .parse, there are also .parse_cron and .parse_duration, for finer granularity.

There is an interesting helper method named .to_duration_hash:

require 'rufus-scheduler'

Rufus::Scheduler.to_duration_hash(60)
  # => { :m => 1 }
Rufus::Scheduler.to_duration_hash(62.127)
  # => { :m => 1, :s => 2, :ms => 127 }

Rufus::Scheduler.to_duration_hash(62.127, drop_seconds: true)
  # => { :m => 1 }

cronline notations specific to rufus-scheduler

first Monday, last Sunday et al

To schedule something at noon every first Monday of the month:

scheduler.cron('00 12 * * mon#1') do
  # ...
end

To schedule something at noon the last Sunday of every month:

scheduler.cron('00 12 * * sun#-1') do
  # ...
end
#
# OR
#
scheduler.cron('00 12 * * sun#L') do
  # ...
end

Such cronlines can be tested with scripts like:

require 'rufus-scheduler'

Time.now
  # => 2013-10-26 07:07:08 +0900
Rufus::Scheduler.parse('* * * * mon#1').next_time.to_s
  # => 2013-11-04 00:00:00 +0900

L (last day of month)

L can be used in the "day" slot:

In this example, the cronline is supposed to trigger every last day of the month at noon:

require 'rufus-scheduler'
Time.now
  # => 2013-10-26 07:22:09 +0900
Rufus::Scheduler.parse('00 12 L * *').next_time.to_s
  # => 2013-10-31 12:00:00 +0900

negative day (x days before the end of the month)

It's OK to pass negative values in the "day" slot:

scheduler.cron '0 0 -5 * *' do
  # do it at 00h00 5 days before the end of the month...
end

Negative ranges (-10--5-: 10 days before the end of the month to 5 days before the end of the month) are OK, but mixed positive / negative ranges will raise an ArgumentError.

Negative ranges with increments (-10---2/2) are accepted as well.

Descending day ranges are not accepted (10-8 or -8--10 for example).

a note about timezones

Cron schedules and at schedules support the specification of a timezone.

scheduler.cron '0 22 * * 1-5 America/Chicago' do
  # the job...
end

scheduler.at '2013-12-12 14:00 Pacific/Samoa' do
  puts "it's tea time!"
end

# or even

Rufus::Scheduler.parse("2013-12-12 14:00 Pacific/Saipan")
  # => #<Rufus::Scheduler::ZoTime:0x007fb424abf4e8 @seconds=1386820800.0, @zone=#<TZInfo::DataTimezone: Pacific/Saipan>, @time=nil>

I get "zotime.rb:41:in `initialize': cannot determine timezone from nil"

For when you see an error like:

rufus-scheduler/lib/rufus/scheduler/zotime.rb:41:
  in `initialize':
    cannot determine timezone from nil (etz:nil,tnz:"中国标准时间",tzid:nil)
      (ArgumentError)
    from rufus-scheduler/lib/rufus/scheduler/zotime.rb:198:in `new'
    from rufus-scheduler/lib/rufus/scheduler/zotime.rb:198:in `now'
    from rufus-scheduler/lib/rufus/scheduler.rb:561:in `start'
    ...

It may happen on Windows or on systems that poorly hint to Ruby which timezone to use. It should be solved by setting explicitly the ENV['TZ'] before the scheduler instantiation:

ENV['TZ'] = 'Asia/Shanghai'
scheduler = Rufus::Scheduler.new
scheduler.every '2s' do
  puts "#{Time.now} Hello #{ENV['TZ']}!"
end

On Rails you might want to try with:

ENV['TZ'] = Time.zone.name # Rails only
scheduler = Rufus::Scheduler.new
scheduler.every '2s' do
  puts "#{Time.now} Hello #{ENV['TZ']}!"
end

(Hat tip to Alexander in gh-230)

Rails sets its timezone under config/application.rb.

Rufus-Scheduler 3.3.3 detects the presence of Rails and uses its timezone setting (tested with Rails 4), so setting ENV['TZ'] should not be necessary.

The value can be determined thanks to https://en.wikipedia.org/wiki/List_of_tz_database_time_zones.

Use a "continent/city" identifier (for example "Asia/Shanghai"). Do not use an abbreviation (not "CST") and do not use a local time zone name (not "中国标准时间" nor "Eastern Standard Time" which, for instance, points to a time zone in America and to another one in Australia...).

If the error persists (and especially on Windows), try to add the tzinfo-data to your Gemfile, as in:

gem 'tzinfo-data'

or by manually requiring it before requiring rufus-scheduler (if you don't use Bundler):

require 'tzinfo/data'
require 'rufus-scheduler'

so Rails?

Yes, I know, all of the above is boring and you're only looking for a snippet to paste in your Ruby-on-Rails application to schedule...

Here is an example initializer:

#
# config/initializers/scheduler.rb

require 'rufus-scheduler'

# Let's use the rufus-scheduler singleton
#
s = Rufus::Scheduler.singleton


# Stupid recurrent task...
#
s.every '1m' do

  Rails.logger.info "hello, it's #{Time.now}"
  Rails.logger.flush
end

And now you tell me that this is good, but you want to schedule stuff from your controller.

Maybe:

class ScheController < ApplicationController

  # GET /sche/
  #
  def index

    job_id =
      Rufus::Scheduler.singleton.in '5s' do
        Rails.logger.info "time flies, it's now #{Time.now}"
      end

    render text: "scheduled job #{job_id}"
  end
end

The rufus-scheduler singleton is instantiated in the config/initializers/scheduler.rb file, it's then available throughout the webapp via Rufus::Scheduler.singleton.

Warning: this works well with single-process Ruby servers like Webrick and Thin. Using rufus-scheduler with Passenger or Unicorn requires a bit more knowledge and tuning, gently provided by a bit of googling and reading, see Faq above.

avoid scheduling when running the Ruby on Rails console

(Written in reply to gh-186)

If you don't want rufus-scheduler to trigger anything while running the Ruby on Rails console, running for tests/specs, or running from a Rake task, you can insert a conditional return statement before jobs are added to the scheduler instance:

#
# config/initializers/scheduler.rb

require 'rufus-scheduler'

return if defined?(Rails::Console) || Rails.env.test? || File.split($PROGRAM_NAME).last == 'rake'
  #
  # do not schedule when Rails is run from its console, for a test/spec, or
  # from a Rake task

# return if $PROGRAM_NAME.include?('spring')
  #
  # see https://github.com/jmettraux/rufus-scheduler/issues/186

s = Rufus::Scheduler.singleton

s.every '1m' do
  Rails.logger.info "hello, it's #{Time.now}"
  Rails.logger.flush
end

(Beware later version of Rails where Spring takes care pre-running the initializers. Running spring stop or disabling Spring might be necessary in some cases to see changes to initializers being taken into account.)

rails server -d

(Written in reply to https://github.com/jmettraux/rufus-scheduler/issues/165 )

There is the handy rails server -d that starts a development Rails as a daemon. The annoying thing is that the scheduler as seen above is started in the main process that then gets forked and daemonized. The rufus-scheduler thread (and any other thread) gets lost, no scheduling happens.

I avoid running -d in development mode and bother about daemonizing only for production deployment.

These are two well crafted articles on process daemonization, please read them:

If, anyway, you need something like rails server -d, why not try bundle exec unicorn -D instead? In my (limited) experience, it worked out of the box (well, had to add gem 'unicorn' to Gemfile first).

executor / reloader

You might benefit from wraping your scheduled code in the executor or reloader. Read more here: https://guides.rubyonrails.org/threading_and_code_execution.html

support

see getting help above.


Author: jmettraux
Source code: https://github.com/jmettraux/rufus-scheduler
License: MIT license

#ruby 

Query of MongoDB | MongoDB Command | MongoDB | Asp.Net Core Mvc

https://youtu.be/FwUobnB5pv8

#mongodb tutorial #mongodb tutorial for beginners #mongodb database #mongodb with c# #mongodb with asp.net core #mongodb

Install MongoDB Database | MongoDB | Asp.Net Core Mvc

#MongoDB
#Aspdotnetexplorer

https://youtu.be/cnwNWzcw3NM

#mongodb #mongodb database #mongodb with c# #mongodb with asp.net core #mongodb tutorial for beginners #mongodb tutorial

Cron Job Scheduling In Laravel

Today I will show you Cron Job Scheduling In Laravel, many time we require to run some piece of code specific interval time period in laravel and we need to run manually every time but command scheduler through we can run and create cron job in laravel.

So, here i will teach you how to create cron job in laravel, and how to create custom command in laravel.

Cron Job Scheduling In Laravel

https://websolutionstuff.com/post/cron-job-scheduling-in-laravel

#cron job scheduling in laravel #laravel #scheduling #scheduler #cron #how to create cron job in laravel

Insert & Save method in MongoDB | Inserting Document | CRUD Operation in MongoDB

#MongoDB
#AspDotNetExplorer
https://youtu.be/CohnNdE_rjM

#mongodb #mongodb tutorial for beginners #mongodb tutorial #mongodb database #learn mongodb