rq-scheduler 0.13.1

Scheduling a Job

There are two ways you can schedule a job. The first is using RQ Scheduler’s enqueue_at

from redis import Redis
from rq import Queue
from rq_scheduler import Scheduler
from datetime import datetime

scheduler = Scheduler(connection=Redis()) # Get a scheduler for the "default" queue
scheduler = Scheduler('foo', connection=Redis()) # Get a scheduler for the "foo" queue

# You can also instantiate a Scheduler using an RQ Queue
queue = Queue('bar', connection=Redis())
scheduler = Scheduler(queue=queue, connection=queue.connection)

# Puts a job into the scheduler. The API is similar to RQ except that it
# takes a datetime object as first argument. So for example to schedule a
# job to run on Jan 1st 2020 we do:
scheduler.enqueue_at(datetime(2020, 1, 1), func) # Date time should be in UTC

# Here's another example scheduling a job to run at a specific date and time (in UTC),
# complete with args and kwargs.
scheduler.enqueue_at(datetime(2020, 1, 1, 3, 4), func, foo, bar=baz)

# You can choose the queue type where jobs will be enqueued by passing the name of the type to the scheduler
# used to enqueue
scheduler = Scheduler('foo', queue_class="rq.Queue")
scheduler.enqueue_at(datetime(2020, 1, 1), func) # The job will be enqueued at the queue named "foo" using the queue type "rq.Queue"

The second way is using enqueue_in. Instead of taking a datetime object, this method expects a timedelta and schedules the job to run at X seconds/minutes/hours/days/weeks later. For example, if we want to monitor how popular a tweet is a few times during the course of the day, we could do something like

from datetime import timedelta

# Schedule a job to run 10 minutes, 1 hour and 1 day later
scheduler.enqueue_in(timedelta(minutes=10), count_retweets, tweet_id)
scheduler.enqueue_in(timedelta(hours=1), count_retweets, tweet_id)
scheduler.enqueue_in(timedelta(days=1), count_retweets, tweet_id)

IMPORTANT: You should always use UTC datetime when working with RQ Scheduler.

Periodic & Repeated Jobs

As of version 0.3, RQ Scheduler also supports creating periodic and repeated jobs. You can do this via the schedule method. Note that this feature needs RQ >= 0.3.1.

This is how you do it

scheduler.schedule(
    scheduled_time=datetime.utcnow(), # Time for first execution, in UTC timezone
    func=func,                     # Function to be queued
    args=[arg1, arg2],             # Arguments passed into function when executed
    kwargs={'foo': 'bar'},         # Keyword arguments passed into function when executed
    interval=60,                   # Time before the function is called again, in seconds
    repeat=10,                     # Repeat this number of times (None means repeat forever)
    meta={'foo': 'bar'}            # Arbitrary pickleable data on the job itself
)

IMPORTANT NOTE: If you set up a repeated job, you must make sure that you either do not set a result_ttl value or you set a value larger than the interval. Otherwise, the entry with the job details will expire and the job will not get re-scheduled.

Cron Jobs

As of version 0.6.0, RQ Scheduler also supports creating Cron Jobs, which you can use for repeated jobs to run periodically at fixed times, dates or intervals, for more info check https://en.wikipedia.org/wiki/Cron. You can do this via the cron method.

This is how you do it

scheduler.cron(
    cron_string,                # A cron string (e.g. "0 0 * * 0")
    func=func,                  # Function to be queued
    args=[arg1, arg2],          # Arguments passed into function when executed
    kwargs={'foo': 'bar'},      # Keyword arguments passed into function when executed
    repeat=10,                  # Repeat this number of times (None means repeat forever)
    result_ttl=300              # Specify how long (in seconds) successful jobs and their results are kept. Defaults to -1 (forever)
    ttl=200                     # Specifies the maximum queued time (in seconds) before it's discarded. Defaults to None (infinite TTL).
    queue_name=queue_name,      # In which queue the job should be put in
    meta={'foo': 'bar'},        # Arbitrary pickleable data on the job itself
    use_local_timezone=False    # Interpret hours in the local timezone
)

Retrieving scheduled jobs

Sometimes you need to know which jobs have already been scheduled. You can get a list of enqueued jobs with the get_jobs method

list_of_job_instances = scheduler.get_jobs()

In it’s simplest form (as seen in the above example) this method returns a list of all job instances that are currently scheduled for execution.

Additionally the method takes two optional keyword arguments until and with_times. The first one specifies up to which point in time scheduled jobs should be returned. It can be given as either a datetime / timedelta instance or an integer denoting the number of seconds since epoch (1970-01-01 00:00:00). The second argument is a boolean that determines whether the scheduled execution time should be returned along with the job instances.

Example

# get all jobs until 2012-11-30 10:00:00
list_of_job_instances = scheduler.get_jobs(until=datetime(2012, 10, 30, 10))

# get all jobs for the next hour
list_of_job_instances = scheduler.get_jobs(until=timedelta(hours=1))

# get all jobs with execution times
jobs_and_times = scheduler.get_jobs(with_times=True)
# returns a list of tuples:
# [(<rq.job.Job object at 0x123456789>, datetime.datetime(2012, 11, 25, 12, 30)), ...]

Checking if a job is scheduled

You can check whether a specific job instance or job id is scheduled for execution using the familiar python in operator

if job_instance in scheduler:
    # Do something
# or
if job_id in scheduler:
    # Do something

Canceling a job

To cancel a job, simply pass a Job or a job id to scheduler.cancel

scheduler.cancel(job)

Note that this method returns None whether the specified job was found or not.

Running the scheduler

RQ Scheduler comes with a script rqscheduler that runs a scheduler process that polls Redis once every minute and move scheduled jobs to the relevant queues when they need to be executed

# This runs a scheduler process using the default Redis connection
rqscheduler

If you want to use a different Redis server you could also do

rqscheduler --host localhost --port 6379 --db 0

The script accepts these arguments:

• -H or --host: Redis server to connect to

• -p or --port: port to connect to

• -d or --db: Redis db to use

• -P or --password: password to connect to Redis

• -b or --burst: runs in burst mode (enqueue scheduled jobs whose execution time is in the past and quit)

• -i INTERVAL or --interval INTERVAL: How often the scheduler checks for new jobs to add to the queue (in seconds, can be floating-point for more precision).

• -j or --job-class: specify custom job class for rq to use (python module.Class)

• -q or --queue-class: specify custom queue class for rq to use (python module.Class)

The arguments pull default values from environment variables with the same names but with a prefix of RQ_REDIS_.

[Unit]
Description=RQScheduler
After=network.target

[Service]
ExecStart=/home/<<User>>/.virtualenvs/<<YourVirtualEnv>>/bin/python \
    /home/<<User>>/.virtualenvs/<<YourVirtualEnv>>/lib/<<YourPythonVersion>>/site-packages/rq_scheduler/scripts/rqscheduler.py

[Install]
WantedBy=multi-user.target

sudo systemctl start rqscheduler.service
sudo systemctl status rqscheduler.service
sudo systemctl enable rqscheduler.service

Running Multiple Schedulers

Multiple instances of the rq-scheduler can be run simultaneously. It allows for

• Reliability (no single point of failure)

• Failover (scheduler instances automatically retry to attain lock and schedule jobs)

• Running scheduler on multiple server instances to make deployment identical and easier

Multiple schedulers can be run in any way you want. Typically you’ll only want to run one scheduler per server/instance.

rqscheduler -i 5

# another shell/systemd service or ideally another server
rqscheduler -i 5

# different parameters can be provided to different schedulers
rqscheduler -i 10

Practical example:

• scheduler_a is running on ec2_instance_a

• If scheduler_a crashes or ec2_instance_a goes down, then our tasks won’t be scheduled at all

• Instead we can simply run 2 schedulers. Another scheduler called scheduler_b can be run on ec2_instance_b

• Now both scheduler_a and scheduler_b will periodically check and schedule the jobs

• If one fails, the other still works

You can read more about multiple schedulers in #212 and #195