raquel

raquel

Simple and elegant Job Queues for Python using SQL.

Tired of complex job queues for distributed computing or event-based systems? Do you want full visibility and complete reliability of your job queue? Raquel is a perfect solution for a distributed task queue and background workers.

• Simple: Use any existing or standalone SQL database. Requires a single table!
• Flexible: Schedule whatever you want however you want. No frameworks, no restrictions.
• Reliable: Uses SQL transactions and handles exceptions, retries, and "at least once" execution. SQL guarantees persistent jobs.
• Transparent: Full visibility into which jobs are running, which failed and why, which are pending, etc. Query anything using SQL.

Installation

pip install raquel

To install with async support, specify the asyncio extra. This simply adds the greenlet package as a dependency.

pip install raquel[asyncio]

Scheduling jobs

In Raquel, jobs are scheduled using the enqueue() method. By default, the job is scheduled for immediate execution. But you can also schedule the job for a specific time in the future or past using the at parameter.

Use the delay parameter to schedule the job to run after a certain amount of time in addition to the at time (yes, you use either or both). The workers won't pick up the job until the scheduled time is reached.

Payload can be any JSON-serializable object or simply a string. It can even be empty. In database, the payload is stored as text for maximum compatibility with all SQL databases, so anything that can be serialized to text can be used as a payload.

If you omit the queue name, the job will be placed into the "default" queue. Use the queue name to place jobs into different queues.

from raquel import Raquel

# Raquel uses SQLAlchemy to connect to most SQL databases
rq = Raquel("postgresql+psycopg2://postgres:postgres@localhost/postgres")
# Enqueing a job is as simple as this
rq.enqueue(queue="messages", payload="Hello, World!")
rq.enqueue(queue="tasks", payload={"data": [1, 2]})

ℹ️ Everything in Raquel is designed to work with both sync and async code. You can use the AsyncRaquel class to enqueue and dequeue jobs in an async manner.

Just don't forget the asyncio extra when installing the package: raquel[asyncio].

import asyncio
from raquel import AsyncRaquel

rq = AsyncRaquel("postgresql+asyncpg://postgres:postgres@localhost/postgres")

async def main():
    await rq.enqueue("default", {'my': {'name_is': 'Slim Shady'}})

asyncio.run(main())

We can also schedule jobs using plain SQL by simply inserting a row into the jobs table. For example, in PostgreSQL:

-- Schedule 3 jobs in the "my-jobs" queue for immediate processing
INSERT INTO jobs 
    (id, queue, status, payload)
VALUES
    (uuid_generate_v4(), 'my-jobs', 'queued', '{"my": "payload"}'),
    (uuid_generate_v4(), 'my-jobs', 'queued', '101'),
    (uuid_generate_v4(), 'my-jobs', 'queued', 'Is this the real life?');

Picking up jobs

Jobs can be picked up from the queues in two main ways.

• The subscribe() decorator
• The dequeue() context manager

The subscribe() decorator

Decorates a function to subscribe to certain queues. This is the simplest approach, because it handles everything for you.

This approach is recommended for most use cases. It works exactly like the dequeue() context manager described next, because it's essentially a while-loop wrapper around it.

The subscribe() approach will run an infinite loop for you, picking up jobs from the queues as they arrive.

Subscribe to queues

from raquel import Raquel, Job

rq = Raquel("postgresql+psycopg2://postgres:postgres@localhost/postgres")

@rq.subscribe("messages", "tasks")
def worker(job: Job):
    do_work(job.payload)

# This loop will run forever, picking up jobs from the "messages" and "tasks"
# queues every second as their scheduled time approaches.
worker.run()

In async mode, the subscribe() decorator works the same way:

import asyncio
from raquel import AsyncRaquel, Job

rq = AsyncRaquel("postgresql+asyncpg://postgres:postgres@localhost/postgres")

@rq.subscribe("messages", "tasks")
async def worker(job: Job):
    await do_work(job.payload)

asyncio.run(worker.run())

Stop the subscription

You can stop the subscription by raising the StopSubscription exception from within the decorated function:

import itertools
counter = itertools.count()

@rq.subscribe("tasks")
def worker(job: Job):
    do_work(job.payload)

    # Stop the subscription after the first 10 jobs are processed
    if next(counter) > 10:
        raise StopSubscription

The dequeue() context manager

This is what does 99% of work behind the subscribe() decorator. Using it directly offers a bit more flexibility but you have to make a call to the dequeue() method each time you want to process a new job.

The important thing to note about the dequeue() is that it's a context manager that yields a Job object for you to work with. It does so by making three consecutive and independent SQL transactions:

• Transaction 1 (optional): Looks for jobs whose max_age is not null and whose enqueued_at + max_age is in the past and updates their status to expired.

• Transaction 2: Selects the next job from the queue and sets its status to claimed, all in one go. It either succeeds in claiming the job or not.

• Transaction 3: Places a database lock on that "claimed" row with the job details for the entire duration of the processing and then updates the job with an appropriate status value:

  • success if the job is processed successfully and we are done with it.
  • failed if an exception was caught by the context manager or the job was manually marked as failed. The job will be rescheduled for a retry.
  • queued if the job was manually rejected or manually rescheduled for a later time.
  • cancelled if the job is manually cancelled.
  • exhausted if the job has reached the maximum number of retries.

All of that happens inside the context manager itself.

while True:
    # dequeue() is a context manager that yields a job object. It will
    # handle the job status and exceptions for you.
    with rq.dequeue("tasks") as job:
        if not job:
            time.sleep(1)
            continue
        do_work(job.payload)

The jobs table

Raquel uses a single database table called jobs. This is all it needs. Can you believe it?

Here is what this table consists of:

Column	Type	Description	Default	Nullable
id	UUID	Unique identifier of the job.		No
queue	TEXT	Name of the queue.	"default"	No
payload	TEXT	Payload of the job. It can be anything. Just needs to be serializable to text.	Null	Yes
status	TEXT	Status of the job.	"queued"	No
max_age	INTEGER	Maximum age of the job in milliseconds.	Null	Yes
max_retry_count	INTEGER	Maximum number of retries.	Null	Yes
min_retry_delay	INTEGER	Minimum delay between retries in milliseconds.	1000	Yes
max_retry_delay	INTEGER	Maximum delay between retries in milliseconds.	12 * 3600 * 1000	Yes
backoff_base	INTEGER	Base in milliseconds for exponential retry backoff.	1000	Yes
enqueued_at	BIGINT	Time when the job was enqueued in milliseconds since epoch (UTC).	now	No
scheduled_at	BIGINT	Time when the job is scheduled to run in milliseconds since epoch (UTC).	now	No
attempts	INTEGER	Number of attempts to execute the job.	0	No
error	TEXT	Error message if the job failed.	Null	Yes
error_trace	TEXT	Error traceback if the job failed.	Null	Yes
claimed_by	TEXT	ID or name of the worker that claimed the job.	Null	Yes
claimed_at	BIGINT	Time when the job was claimed in milliseconds since epoch (UTC).	Null	Yes
finished_at	BIGINT	Time when the job was finished in milliseconds since epoch (UTC).	Null	Yes

Job status

Jobs can have the following statuses:

• queued - Job is waiting to be picked up by a worker.
• claimed - Job is currently locked and is being processed by a worker (in databases such as PostgreSQL, MySQL, etc., once the job is claimed, its row is locked until the worker is done with it).
• success - Job was successfully executed.
• failed - Job failed to execute. This happens when an exception was caught by the dequeue() context manager. The last error message and traceback are stored in the error and error_trace columns. Job will be retried again, meaning it will be rescheduled with an exponential backoff delay.
• cancelled - Job was manually cancelled.
• expired - Job was not picked up by a worker in time (when max_age is set).
• exhausted - Job has reached the maximum number of retries (when max_retry_count is set).

The job can be picked up by a worker in either of the following three states: queued, failed, or claimed.

The first two are most common. The job will be picked up if:

• Job status is queued (scheduled or rejected) or failed (failed to be processed and is being retried);
• And its scheduled_at time is in the past;
• And its max_age is not set or its enqueued_at + max_age is in the future.

The job in claimed state is a special case and happens when some worker marked the job as claimed but failed to process it. In this case the row, representing the job, is not locked in the database. If more than a minute passed since the job was claimed (claimed_at + 1 minute), any worker can reclaim it for itself.

How do we guarantee that the same job is not picked up by multiple workers?

Short answer: by locking the row and using the claimed status.

In PostgreSQL, we use the SELECT FOR UPDATE SKIP LOCKED statement is used. In other databases that support this (such as Oracle, MySQL) a similar approach is used.

In extremely simple databases, such as SQLite, the fact that the whole database is locked during a write operation guarantees that no other worker will be able to set the job status to claimed at the same time.

What happens if a worker dies?

• If a worker dies while attempting to claim a job, the transaction opened by the worker is rolled back and the row is unlocked by the database. Another worker can claim it and process it.
• If a worker dies while processing a job, the row is unlocked by the database but remains in the claimed status. Another worker can pick it up and process it.

Fancy ways to schedule jobs

All jobs are scheduled to run at a specific time. By default, this time is set to the current time when the job is enqueued. You can set the scheduled_at field to a future time to schedule the job to run at that time.

rq.enqueue("my-jobs", 10_000, at=datetime.now() + timedelta(hours=10))

But you can also specify a delay to schedule the job to run in the future:

# Same as above
rq.enqueue("my-jobs", 10_000, delay=timedelta(hours=10))

When enqueuing a job, you can specify all the parameters that are available in the jobs table. For example:

• max_age - Maximum age of the job in milliseconds.
• max_retry_count - Maximum allowed number of retries.
• min_retry_delay - Minimum delay between retries in milliseconds.
• max_retry_delay - Maximum delay between retries in milliseconds.
• backoff_base - Base in milliseconds for exponential retry backoff.

Reschedule a job

The reschedule() method is used to reprocess the job at a later time. The job will remain in the queue with a new scheduled execution time, and the current attempt won't count towards the maximum number of retries.

This method should only be called inside the dequeue() context manager or the subscribe() decorator.

@rq.subscribe("my-queue")
def worker(job: Job):
    # Check if we have everything ready to process the job, and if not,
    # reschedule the job to run 10 minutes from now
    if not is_everything_ready_to_process(job.payload):
        job.reschedule(delay=timedelta(minutes=10))
    else:
        # Otherwise, process the job
        do_work(job.payload)

When you reschedule a job, its scheduled_at field is either updated with the new at and delay values or left unchanged. And the finished_at field is cleared. If the Job object had any error or error_trace values, they are saved to the database. The attempts field is incremented.

You can find all rescheduled jobs using SQL by filtering for those that are queued, but have attempts and were claimed before.

SELECT * FROM jobs
WHERE status = 'queued'
AND attempts > 0
AND claimed_at IS NOT NULL

Here are some fancy ways to reschedule a job using the same reschedule() method:

ℹ️ Make sure to call reschedule() inside the dequeue() context manager or the subscribe() decorator.

# Run when the next day starts
job.reschedule(
    at=datetime.now().replace(
        hour=0, minute=0, second=0, microsecond=0,
    ) + timedelta(days=1)
)

# Same but using the `delay` parameter
job.reschedule(
    at=datetime.now().replace(hour=0, minute=0, second=0, microsecond=0),
    delta=timedelta(days=1),
)

# Run in 500 milliseconds
job.reschedule(delay=500)

# Run in `min_retry_delay` milliseconds, as configured for this job
# (default is 1 second)
job.reschedule()

Reject the job

In case your worker can't process the job for some reason, you can reject it, allowing it to be immediately claimed by another worker.

This method should only be called inside the dequeue() context manager or the subscribe() decorator.

It is very similar to rescheduling the job to run immediately. When you reject the job, the scheduled_at field is left unchanged, but the claimed_at and claimed_by fields are cleared. The job status is set to queued. And the attempts field is incremented.

@rq.subscribe("my-queue")
def worker(job: Job):
    # Maybe this worker doesn't have the necessary permissions to process
    # this job.
    if job.payload.get("requires_admin"):
        job.reject()
        return

    # Otherwise, process the job
    do_work(job.payload)

Here is how you can find all rejected jobs using SQL:

SELECT * FROM jobs
WHERE status = 'queued'
AND attempts > 0
AND claimed_at IS NULL

Queue names

By default, all jobs are placed into the "default" queue. You can specify the queue name when enqueuing a job:

# Enqueue a payload into the "default" queue
rq.enqueue(payload="Hello, World!")

# Same as above
rq.enqueue("default", "Hello, World!")

# Enqueue a payload into the "my-queue" queue
rq.enqueue("my-queue", "Hello, World!")

By default, when jobs are dequeued by a worker, the job that has the earliest scheduled_at time is picked up from any queue. If you want to dequeue jobs from specific queues, you can specify the queue names as arguments to the subscribe() decorator or the dequeue() context manager:

# Just one job will be picked up from any of the queues.
with rq.dequeue("my-queue", "another-queue", "as-many-as-you-want"):
    do_work(job.payload)

# Continuously picks up jobs from any of the specified queues.
@rq.subscribe("my-queue", "another-queue", "as-many-as-you-want")
def worker(job: Job):
    do_work(job.payload)

worker.run()

Here is how you can find all jobs from a specific queue using SQL:

SELECT * FROM jobs
WHERE queue = 'my-queue'

And this query can show number of jobs per queue:

SELECT queue, COUNT(*) FROM jobs
GROUP BY queue

Job retries

Jobs are retried when they fail. When an exception is caught by the dequeue() context manager, the job is rescheduled with an exponential backoff delay. The same logic applies within the subscribe() decorator.

By default, the job will be retried indefinitely. You can set the max_retry_count or max_age fields to limit the number of retries or the maximum age of the job.

@rq.subscribe()
def worker(job: Job):
    # Let the context manager handle the exception for you.
    # The exception will be caught and the job will be retried.
    # Under the hood, context manager will call `job.fail()` for you.
    raise Exception("Oh no")
    do_work(job.payload)

You can always handle the exception manually:

@rq.subscribe()
def worker(job: Job):
    # Catch an exception manually
    try:
        do_work(job.payload)
    except Exception as e:
        # If you mark the job as failed, it will be retried.
        job.fail(str(e))

Whenever job fails, the error and the traceback are stored in the error and error_trace columns. The job status is set to failed and the job will be retried. The attempt number is incremented.

Here is how you can find all failed jobs using SQL:

SELECT * FROM jobs
WHERE status = 'failed'

Retry delay

The next retry time after a failed job is calculated as follows:

• Take the current scheduled_at time.
• Add the time it took to process the job (aka duration).
• Add the retry delay.

The retry delay itself is calculated as follows:

backoff_base * 2 ^ attempt

But this is just a planned retry delay. The actual retry delay is capped between the min_retry_delay and max_retry_delay values. The min_retry_delay defaults to 1 second and the max_retry_delay defaults to 12 hours. The backoff_base defaults to 1 second.

In other words, here is how your job will be retried (assuming there is always a worker available and the job takes almost no time to process):

Retry	delay
1	1 second after 1st attempt
2	2 seconds after 2nd attempt
3	after 4 seconds
...	...
6	after ~2 minutes
...	...
10	after ~30 minutes
...	...
14	after ~9 hours
...	...

and so on with the maximum delay of 12 hours, or based on the maximum value you set for this job using the max_retry_delay setting.

For certain types of jobs, it makes sense to chill out for a bit before retrying. For example, an API might have a rate limit you've just hit or some data might not be ready yet. In such cases, you can set the min_retry_delay to a higher value, such as 10 or 30 seconds.

ℹ️ Remember, that all durations and timestamps are in milliseconds. So 10 seconds is 10 * 1000 = 10000 milliseconds. 1 minute is 60 * 1000 = 60000 milliseconds.

Preparing your database

Create the jobs table by calling a function

You can configure the table using the create_all() method, which will automatically use the supported syntax for the database you are using (it is safe to run it multiple times, it only creates the table once.).

# Works for all databases
rq.create_all()

Create the table using SQL

Alternatively, you can create the table manually. For example, in Postgres you can use the following SQL:

CREATE TABLE IF NOT EXISTS jobs (
    id UUID PRIMARY KEY DEFAULT uuid_generate_v4(),
    queue TEXT NOT NULL,
    payload TEXT,
    status TEXT NOT NULL DEFAULT 'queued',
    max_age BIGINT,
    max_retry_count INTEGER,
    min_retry_delay INTEGER DEFAULT 1000,
    max_retry_delay INTEGER DEFAULT 43200000,
    backoff_base INTEGER DEFAULT 1000,
    enqueued_at BIGINT NOT NULL DEFAULT extract(epoch from now()) * 1000,
    scheduled_at BIGINT NOT NULL DEFAULT extract(epoch from now()) * 1000,
    attempts INTEGER NOT NULL DEFAULT 0,
    error TEXT,
    error_trace TEXT,
    claimed_by TEXT,
    claimed_at BIGINT,
    finished_at BIGINT
);
CREATE INDEX IF NOT EXISTS idx_jobs_queue ON jobs (queue);
CREATE INDEX IF NOT EXISTS idx_jobs_status ON jobs (status);
CREATE INDEX IF NOT EXISTS idx_jobs_scheduled_at ON jobs (scheduled_at);
CREATE INDEX IF NOT EXISTS idx_jobs_claimed_by ON jobs (claimed_by);

Create the table using Alembic migrations

Import Raquel's metadata and add it as a target inside the context.configure() calls in the env.py file:

# Alembic's env.py configuration file

from raquel.models.base_sql import BaseSQL as RaquelBaseSQL

# This code already exists in the Alembic configuration file
def run_migrations_offline() -> None:
    # ...
    context.configure(
        # ...
        target_metadata=[
            target_metadata,
            RaquelBaseSQL.metadata,  # <- Add Raquel's metadata
        ],
        # ...
    )

# Same for online migrations
def run_migrations_online() -> None:
    # ...
    context.configure(
        # ...
        target_metadata=[
            target_metadata,
            RaquelBaseSQL.metadata,  # <- Add Raquel's metadata
        ],
        # ...
    )

Once you do that, Alembic will automatically add the jobs table and the appropriate indexes to your migration file.

Fun facts

• Raquel is named after the famous actress Raquel Welch, who I have never seen in a movie. But her poster was one of the cutouts hanging in the shrine dedicated to Arnold Schwarzenegger in the local gym I used to attend. Don't ask me why and how.
• The name Raquel is also a combination of the words "queue" and "SQL".
• The payload can be empty or Null (None). You can use it to schedule jobs without any payload, for example, to send a notification or to run a periodic task.

Should I trust Raquel with my production?

Yes, you should! Here is why:

• Raquel is dead simple. Unbelievably so.

  You can rewrite it in any programming language using plain SQL in about a day. We are not going to overcomplicate things here, because everyone is already sick of Celery and Co.

• It's incredibly reliable.

  The jobs are stored in a relational database and exclusive locks and rollbacks are handled through ACID transactions.

• It relies on a single dependency: SQLAlchemy.

  The most mature and well-tested database library for Python. The codebase of Raquel can remain same for ages, but you still get support for all the new databases and bugfixes. We do not pin SQLAlchemy dependency strictly.

• Raquel is already used in production by several companies.

  Examples:

  • Dynatrace

• Raquel is licensed under the Apache 2.0 license.

  You can use it in any project or fork it and do whatever you want with it.

• Raquel is actively maintained by Vagiz Duseev.

  I'm also the author of some other popular Python packages such as opensearch-logger.

• Raquel is as platform and database agnostic.

  Save yourself the pain of migrating between database vendors or versions. All timestamps are stored as milliseconds since epoch (UTC timezone). Payloads are stored as text. Job IDs are random UUIDs to allow migration between databases and HA setups.

Comparison to alternatives

Feature	Raquel	Celery	RQ	Dramatiq	arq	pgqueuer	pq
Special tooling to run workers	No ✅	Yes ❌	Yes ❌	Yes ❌	Yes ❌	Yes ❌	Yes ❌
Needs message queue	No ✅	Yes ❌	Yes ❌	Yes ❌	Yes ❌	Yes ❌	Yes ❌
Supports SQL	Yes ✅	No ❌	No ❌	No ❌	No ❌	Yes ✅	Yes ✅
Full visibility	Yes ✅	No ❌	No ❌	No ❌	No ❌	Yes ✅	Yes ✅
Reliable	Yes ✅	Yes ✅	Yes ✅	Yes ✅	Yes ✅	Yes ✅	Yes ✅
Supports async	Yes ✅	Yes ❌	Yes ❌	Yes ❌	Yes ❌	Yes ❌	Yes ❌
Persistent jobs	Yes ✅	Yes ❌	Yes ❌	Yes ❌	Yes ❌	Yes ✅	Yes ✅
Schedule from anywhere	Yes ✅	No ❌	No ❌	Yes ✅	No ❌	Yes ✅	Yes ✅
Job payload size limit	No ✅	Yes ❌	Yes ❌	Yes ❌	Yes ❌	Yes ❌	Yes ❌

As you can clearly see, not only is Raquel a superior distributed task queue, job queue, and background worker system, but very likely the best software ever written in the history of mankind. And I claim this without any bias.