RedBeat

RedBeat is a Celery Beat Scheduler that stores the scheduled tasks and runtime metadata in Redis.

Why RedBeat

1. Dynamic live task creation and modification, without lengthy downtime
2. Externally manage tasks from any language with Redis bindings
3. Shared data store; Beat isn't tied to a single drive or machine
4. Fast startup even with a large task count
5. Prevent accidentally running multiple Beat servers

Getting Started

Install with pip:

# pip install celery-redbeat

Configure RedBeat settings in your celery configuration file:

REDBEAT_REDIS_URL = "redis://localhost:6379/1"

Then specify the scheduler when running Celery Beat:

$ celery beat -S redbeat.RedBeatScheduler

RedBeat uses a distributed lock to prevent multiple instances running. To disable this feature, set:

REDBEAT_LOCK_KEY = None

Configuration

You can add any of the following parameters to your celery configuration:

REDBEAT_REDIS_URL: URL to redis server used to store the schedule
REDBEAT_KEY_PREFIX: A prefix for all keys created by RedBeat, default 'redbeat'
REDBEAT_LOCK_KEY: Key used to ensure only a single beat instance runs at a time
REDBEAT_LOCK_TIMEOUT: Unless refreshed the lock will expire after this time

Design

At its core RedBeat uses a Sorted Set to store the schedule as a priority queue. It stores task details using a hash key with the task definition and metadata.

The schedule set contains the task keys sorted by the next scheduled run time.

For each tick of Beat

1. get list of due keys and due next tick
2. retrieve definitions and metadata for all keys from previous step
3. update task metadata and reschedule with next run time of task
4. call due tasks using async_apply
5. calculate time to sleep until start of next tick using remaining tasks

Creating Tasks

You can use the standard CELERYBEAT_SCHEDULE to define static tasks or you can insert tasks directly into Redis.

The easiest way to insert tasks from Python is it use `RedBeatSchedulerEntry()`:

interval = celey.schedulers.schdule(run_every=60)  # seconds
entry = RedBeatSchedulerEntry('task-name', 'tasks.some_task', interval, args=['arg1', 2])
entry.save()

Alternatively, you can insert directly into Redis by creating a new hash with a key of REDBEAT_KEY_PREFIX:task-name. It should contain a single key definition which is a JSON blob with the task details.

Interval

An interval task is defined with the JSON like:

{
    "name" : "interval example",
    "task" : "tasks.every_5_seconds",
    "schedule": {
        "__type__": "interval",
        "every" : 5, # seconds
        "relative": false, # optional
    },
    "args" : [  # optional
        "param1",
        "param2"
    ],
    "kwargs" : {  # optional
        "max_targets" : 100
    },
    "enabled" : true,  # optional
}

Crontab

An crontab task is defined with the JSON like:

{
    "name" : "crontab example",
    "task" : "tasks.daily",
    "schedule": {
        "__type__": "crontab",
        "minute" : "5", # optional, defaults to *
        "hour" : "*", # optional, defaults to *
        "day_of_week" : "monday", # optional, defaults to *
        "day_of_month" : "*/7", # optional, defaults to *
        "month_of_year" : "[1-12]", # optional, defaults to *
    },
    "args" : [  # optional
        "param1",
        "param2"
    ],
    "kwargs" : {  # optional
        "max_targets" : 100
    },
    "enabled" : true,  # optional
}

Scheduling

You will also need to insert the new task into the schedule with:

zadd REDBEAT_KEY_PREFIX::schedule 0 new-task-name

The score is the next time the task should run formatted as a UNIX timestamp.

Metadata

Applications may also want to manipulate the task metadata to have more control over when a task runs. The meta key contains a JSON blob as follows:

{
    'last_run_at': {
        '__type__': 'datetime',
        'year': 2015,
        'month': 12,
        'day': 29,
        'hour': 16,
        'minute': 45,
        'microsecond': 231
    },
    'total_run_count'; 23
}

For instance by default `last_run_at` corresponds to when Beat dispatched the task, but depending on queue latency it might not run immediately, but the application could update the metadata with the actual run time, allowing intervals to be relative to last execution rather than last dispatch.

Development

RedBeat is available on GitHub

Once you have the source you can run the tests with the following commands:

pip install -r requirements.dev.txt
py.test tests

You can also quickly fire up a sample Beat instance with:

celery beat --config exampleconf