Revel provides a framework for performing work asynchronously, outside of the request flow. This may take the form of recurring tasks that update cached data or one-off tasks for sending emails.

Activation

The framework is included as an optional module, that is not included in your application by default. To activate it, add the module to your app configuration:

module.jobs = github.com/robfig/revel/modules/jobs

Additionally, in order to access the job monitoring page, you will need to add this line to your routes:

module:jobs

This statement will insert the /@jobs route at that location.

Options

There are a couple of options that tell the framework what sort of limitations to place on the jobs that it runs.

This example shows them set to their default values.

jobs.pool = 10                # Number of jobs allowed to run simultaneously
jobs.selfconcurrent = false   # Allow a job to run only if previous instances are done

Startup jobs

To run a task on application startup, use revel.OnAppStart to register a function. Revel runs these tasks serially, before starting the server. Note that this functionality does not actually use the jobs module, but it can be used to submit a job for execution that doesn’t block server startup.

func init() {
    revel.OnAppStart(func() { jobs.Now(populateCache{}) })
}

Recurring jobs

Jobs may be scheduled to run on any schedule. There are two options for expressing the schedule:

  1. A cron specification
  2. A fixed interval

Revel uses the cron library to parse the schedule and run the jobs. The library’s README provides a detailed description of the format accepted.

Jobs are generally registered using the revel.OnAppStart hook, but they may be registered at any later time as well.

Here are some examples:

import (
    "github.com/robfig/revel"
    "github.com/robfig/revel/modules/jobs/app/jobs"
    "time"
)

type ReminderEmails struct {
    // filtered
}

func (e ReminderEmails) Run() {
    // Queries the DB
    // Sends some email
}

func init() {
    revel.OnAppStart(func() {
        jobs.Schedule("0 0 0 * * ?",  ReminderEmails{})
        jobs.Schedule("@midnight",    ReminderEmails{})
        jobs.Schedule("@every 24h",   ReminderEmails{})
        jobs.Every(24 * time.Hour,    ReminderEmails{})
    })
}

Named schedules

You can configure schedules in your app.conf file and reference them anywhere. This can help provide easy reuse and a useful description for crontab specs.

To define your named cron schedule, put this in your app.conf:

cron.workhours_15m = 0 */15 9-17 ? * MON-FRI

Use the named schedule by referencing it anywhere you would have used a cron spec.

func init() {
    revel.OnAppStart(func() {
        jobs.Schedule("cron.workhours_15m", ReminderEmails{})
    })
}

Note: Your cron schedule name must begin with “cron.”

One-off jobs

Sometimes you will want to do something in response to a user action. In these cases, the jobs module allows you to submit a job to be run a single time.

The only control offered is how long to wait until the job should be run.

type AppController struct { *revel.Controller }

func (c AppController) Action() revel.Result {
    // Handle the request.
    ...

    // Send them email asynchronously, right now.
    jobs.Now(SendConfirmationEmail{})

    // Or, send them email asynchronously after a minute.
    jobs.In(time.Minute, SendConfirmationEmail{})
}

Registering functions

It is possible to register a func() as a job by wrapping it in the jobs.Func type. For example:

func sendReminderEmails() {
    // Query the DB
    // Send some email
}

func init() {
    revel.OnAppStart(func() {
        jobs.Schedule("@midnight", jobs.Func(sendReminderEmails))
    })
}

Job Status

The jobs module provides a status page that shows a list of the scheduled jobs it knows about, their current status (IDLE or RUNNING), and their previous and next run times.

Job Status Page

For security purposes, the status page is restricted to requests that originate from 127.0.0.1.

Constrained pool size

It is possible to configure the job module to limit the number of jobs that are allowed to run at the same time. This allows the developer to restrict the resources that could be potentially in use by asynchronous jobs – typically interactive responsiveness if valued above asynchronous processing. When a pool is full of running jobs, new jobs block to wait for running jobs to complete.

Implementation note: The implementation blocks on a channel receive, which is implemented to be FIFO for waiting goroutines (but not specified/required to be so). See here for discussion.

Areas for development

  • Allow access to the job status page with HTTP Basic Authentication credentials
  • Allow administrators to run scheduled jobs interactively from the status page
  • Provide more visibility into the job runner, e.g. the pool size, the job queue length, etc.