plain.jobs

Process background jobs with a database-driven job queue.

• Overview
• Local development
• Job parameters
• Job methods
• Scheduled jobs
• Admin interface
• Job history
• Worker resilience
• Monitoring
• Settings
• FAQs
• Idempotency
• Installation

Overview

Jobs are defined using the Job base class and the run() method at a minimum.

from plain.jobs import Job, register_job
from plain.email import send_mail


@register_job
class WelcomeUserJob(Job):
    def __init__(self, user):
        self.user = user

    def run(self):
        send_mail(
            subject="Welcome!",
            message=f"Hello from Plain, {self.user}",
            from_email="[email protected]",
            recipient_list=[self.user.email],
        )

You can then create an instance of the job and call run_in_worker() to enqueue it for a background worker to pick up.

user = User.query.get(id=1)
WelcomeUserJob(user).run_in_worker()

Workers are run using the plain jobs worker command.

Jobs can be defined in any Python file, but it is suggested to use app/jobs.py or app/{pkg}/jobs.py as those will be imported automatically so the @register_job decorator will fire.

Sync the database after installation:

plain postgres sync

Local development

In development, you will typically want to run the worker alongside your app with auto-reloading enabled. With plain.dev you can do this by adding it to the [tool.plain.dev.run] section of your pyproject.toml file.

# pyproject.toml
[tool.plain.dev.run]
worker = {cmd = "plain jobs worker --reload --stats-every 0 --max-processes 2"}
worker-slow = {cmd = "plain jobs worker --reload --queue slow --stats-every 0 --max-processes 2"}

The --reload flag will automatically watch .py and .env* files for changes and restart the worker when changes are detected.

Job parameters

When calling run_in_worker(), you can specify several parameters to control job execution:

job.run_in_worker(
    queue="slow",  # Target a specific queue (default: "default")
    delay=60,  # Delay in seconds (or timedelta/datetime)
    priority=10,  # Higher numbers run first (default: 0, use negatives for lower priority)
    retries=3,  # Number of retry attempts (default: 0)
    concurrency_key="user-123-welcome",  # Identifier for grouping/deduplication
)

For more advanced parameter options, see Job.run_in_worker().

Job methods

The Job base class provides several methods you can override to customize behavior:

class MyJob(Job):
    def run(self):
        # Required: The main job logic
        pass

    # Defaults (can be overridden in run_in_worker)
    def default_queue(self) -> str:
        return "default"

    def default_priority(self) -> int:
        # Higher numbers run first: 10 > 5 > 0 > -5 > -10
        return 0

    def default_retries(self) -> int:
        return 0

    def default_concurrency_key(self) -> str:
        # Identifier for grouping/deduplication
        return ""

    # Computed values
    def calculate_retry_delay(self, attempt: int) -> int:
        # Delay in seconds before retry (attempt starts at 1)
        return 0

    # Hooks
    def should_enqueue(self, concurrency_key: str) -> bool:
        # Called before enqueueing - return False to skip
        # Use for concurrency limits, rate limits, etc.
        return True

    def on_aborted(self, result: JobResult) -> None:
        # Called when this job's process was terminated externally before
        # run() could complete (status=LOST or status=CANCELLED).
        # Use this to reconcile domain state that run()'s try/finally would
        # have released, since try/finally did not execute.
        pass

Scheduled jobs

Schedules are configured via the JOBS_SCHEDULE setting as a list of (job, schedule) tuples. The job can be a dotted path to a Job subclass (or a "cmd:<shell command>" string), and the schedule can be a cron expression string or a Schedule instance:

# app/settings.py
JOBS_SCHEDULE = [
    ("app.reports.jobs.DailyReportJob", "0 9 * * *"),  # Every day at 9 AM
    ("cmd:./scripts/cleanup", "@daily"),
]

Cron expressions support standard syntax and special strings:

• @yearly or @annually - Run once a year
• @monthly - Run once a month
• @weekly - Run once a week
• @daily or @midnight - Run once a day
• @hourly - Run once an hour

For custom schedules, see Schedule.

Admin interface

The jobs package includes admin views for monitoring jobs under the "Jobs" section. The admin interface provides:

• Requests: View pending jobs in the queue
• Processes: Monitor currently running jobs
• Results: Review completed and failed job history

Dashboard cards show at-a-glance statistics for successful, errored, lost, and retried jobs.

Job history

Job execution history is stored in the JobResult model. This includes:

• Job class and parameters
• Start and end times
• Success/failure status
• Error messages and tracebacks for failed jobs
• Worker information

See Settings for configuring job retention and timeouts.

Worker resilience

Each worker process registers itself in a WorkerHeartbeat row at startup, bumps last_heartbeat_at every JOBS_HEARTBEAT_INTERVAL seconds while running, and deletes the row on clean shutdown. Every JobProcess is stamped with the picking worker's worker_id, so when a heartbeat goes stale (older than JOBS_HEARTBEAT_TIMEOUT), the next worker's rescue tick can find the dead worker's in-flight jobs and convert them to JobResult(status=LOST).

This means a worker killed mid-run (Heroku deploy SIGKILL, OOM, host crash) gets detected within a few minutes — independent of how long any individual job legitimately runs. A long-running job is safe as long as its worker keeps heartbeating.

Reacting to abortions

Override Job.on_aborted(result) to react when a job's process is terminated externally:

@register_job
class GenerateReportJob(Job):
    def __init__(self, report_id):
        self.report_id = report_id

    def run(self):
        report = Report.query.get(id=self.report_id)
        report.status = "running"
        report.save()
        # ...do the work...
        report.status = "done"
        report.save()

    def on_aborted(self, result):
        # The worker was killed mid-run; run()'s cleanup never executed.
        # Flip the report out of the limbo state.
        Report.query.filter(id=self.report_id).update(
            status="failed", error="worker terminated"
        )

The hook fires for terminal statuses that run() itself could not observe — LOST (process killed) and CANCELLED (future cancelled before execution). It does not fire for SUCCESSFUL or ERRORED, which run() can handle directly via normal control flow.

The Job instance is reconstructed from stored parameters; no in-memory state from the original run is preserved. Exceptions raised by the hook are caught and logged so they cannot block the result from being recorded.

If your job has retries configured, on_aborted fires when the JobProcess transitions to LOST/CANCELLED — before any retry runs. Apps that only want to react after retries are exhausted should check result.retries == result.retry_attempt.

Monitoring

Workers report statistics and can be monitored using the --stats-every option:

# Report stats every 60 seconds
plain jobs worker --stats-every 60

The worker integrates with OpenTelemetry for distributed tracing. Spans are created for:

• Job scheduling (run_in_worker) — emits a send {queue} PRODUCER span with the OTel messaging.* semconv attributes
• Job execution — emits a process {queue} CONSUMER span linked back to the originating send span
• Job completion/failure — recorded as the span's status and error.type attribute on failure

Jobs are linked to the originating trace context, allowing you to follow jobs initiated from web requests.

Messaging metrics:

• messaging.client.sent.messages — counter incremented for each enqueue
• messaging.client.consumed.messages — counter incremented for every terminal JobResult. Carries a plain.jobs.outcome attribute (successful, errored, lost, cancelled, deferred) so dashboards can split throughput by outcome.
• messaging.client.operation.duration — histogram of enqueue/process durations
• plain.jobs.queue.wait.duration — histogram of how long a job waited in queue before a worker picked it up

Per-worker observable gauges (queryable per messaging.destination.name where applicable):

• plain.jobs.worker.processes — OS processes spawned by this worker
• plain.jobs.queue.depth — pending JobRequests ready to run
• plain.jobs.queue.scheduled — JobRequests with start_at in the future
• plain.jobs.queue.oldest.age — age in seconds of the oldest ready-to-run JobRequest
• plain.jobs.running — JobProcess rows currently running

Worker-liveness gauge (global, no per-queue dimension):

• plain.jobs.workers — WorkerHeartbeat row count, split by a plain.jobs.worker.state attribute taking active (within JOBS_HEARTBEAT_TIMEOUT) or stale (past it, eligible for rescue on the next tick)

Two contract details to be aware of:

• Successful enqueues record metrics on transaction commit. If you call run_in_worker inside a transaction that later rolls back, the message was never actually persisted — so the counter and histogram do not fire. This matches the OTel semconv: "MUST NOT count messages that were created but haven't yet been sent." Failed enqueues record immediately so transient errors are still visible.
• Skipped enqueues are visible in spans, not in metrics. When should_enqueue returns False (e.g., a concurrency-key collision), the span gets job.enqueue.skipped=True but no metric is recorded — there was no send to count.
• Observable gauges emit once per worker process. When two workers cover the same queue, the per-queue gauges emit identical values from each. plain.jobs.workers is global (no per-queue dimension) and likewise emits the full table count from every worker. Aggregate these gauges with last_value/max, never sum.

Settings

The JOBS_WORKER_* settings configure the plain jobs worker command and can also be overridden via CLI flags (e.g. --max-processes).

See default_settings.py for more details.

FAQs

How do I ensure only one job runs at a time?

Set a concurrency_key to automatically enforce uniqueness - only one job with the same key can be pending or processing:

from plain.jobs import Job, register_job

@register_job
class ProcessUserJob(Job):
    def __init__(self, user_id):
        self.user_id = user_id

    def default_concurrency_key(self):
        return f"user-{self.user_id}"

    def run(self):
        process_user(self.user_id)

# Usage
ProcessUserJob(123).run_in_worker()  # Enqueued
ProcessUserJob(123).run_in_worker()  # Returns None (blocked - job already pending/processing)

Alternatively, pass concurrency_key as a parameter to run_in_worker() instead of overriding the method.

How do I implement custom concurrency limits?

Use the should_enqueue() hook to implement custom concurrency control:

class ProcessUserDataJob(Job):
    def __init__(self, user_id):
        self.user_id = user_id

    def default_concurrency_key(self):
        return f"user-{self.user_id}"

    def should_enqueue(self, concurrency_key):
        # Only allow 1 job per user at a time
        processing = self.get_processing_jobs(concurrency_key).count()
        pending = self.get_requested_jobs(concurrency_key).count()
        return processing == 0 and pending == 0

For more patterns like rate limiting and global limits, see should_enqueue() in the source code.

How are race conditions prevented?

Plain uses PostgreSQL's advisory locks to ensure should_enqueue() checks are atomic with job creation. The lock is acquired during the transaction and automatically released when the transaction completes. This eliminates race conditions where multiple threads might simultaneously pass the should_enqueue() check.

For custom locking behavior (Redis, etc.), override get_enqueue_lock().

Can I run multiple workers?

Yes, you can run multiple worker processes:

plain jobs worker --max-processes 4

Or run workers for specific queues:

plain jobs worker --queue slow --max-processes 2

How do I handle job failures?

Set the number of retries and implement retry delays:

class MyJob(Job):
    def default_retries(self):
        return 3

    def calculate_retry_delay(self, attempt):
        # Exponential backoff: 1s, 2s, 4s
        return 2 ** (attempt - 1)

Idempotency

Jobs may retry on failure, so design them so re-execution is safe.

# Bad — sends duplicate emails on retry
@register_job
class WelcomeUserJob(Job):
    def __init__(self, user):
        self.user = user

    def run(self):
        send_welcome_email(self.user)

# Good — check before acting
@register_job
class WelcomeUserJob(Job):
    def __init__(self, user):
        self.user = user

    def run(self):
        if not self.user.welcome_email_sent:
            send_welcome_email(self.user)
            self.user.welcome_email_sent = True
            self.user.save()

Installation

Install the plain.jobs package from PyPI:

uv add plain.jobs

Add to your INSTALLED_PACKAGES:

# app/settings.py
INSTALLED_PACKAGES = [
    ...
    "plain.jobs",
]