Python RQ Prometheus Exporter

Prometheus metrics exporter for Python RQ (Redis Queue) job queue library.

Installation

Install the Python package:

$ # Install the latest version
$ pip install rq-exporter
$ # Or install a specific version
$ pip install rq-exporter==2.1.0

Or download the Docker image:

$ # Pull the latest image
$ docker pull mdawar/rq-exporter
$ # Or pull a specific version
$ docker pull mdawar/rq-exporter:v2.1.0

The releases are available as Docker image tags.

Usage

Python package:

$ # Start the exporter on port 9726
$ rq-exporter
$ # Start the exporter on a specific port and host (Default: 0.0.0.0:9726)
$ rq-exporter --host localhost --port 8080
$ # By default the exporter will connect to Redis on `localhost` port `6379`
$ # You can specify a Redis URL
$ rq-exporter --redis-url redis://:123456@redis_host:6379/0
$ # Or specific Redis options (host, port, db, password)
$ rq-exporter --redis-host 192.168.1.10 --redis-port 6380 --redis-pass 123456 --redis-db 1
$ # You can also specify a password file path (eg: mounted Docker secret)
$ rq-exporter --redis-pass-file /run/secrets/redis_pass

Docker image:

$ # Run the exporter and publish the port 9726 on the host
$ docker run -it -p 9726:9726 rq-exporter
$ # Use the -d option to run the container in the background (detached)
$ docker run -d -p 9726:9726 rq-exporter
$ # All the command line arguments will be passed to rq-exporter
$ docker run -it -p 9726:9726 rq-exporter --redis-host redis --redis-pass 123456
$ # You can also configure the exporter using environment variables
$ docker run -it -p 9726:9726 -e RQ_REDIS_HOST=redis -e RQ_REDIS_PASS=123456 rq-exporter

Grafana Dashboard

An example Grafana dashboard is available with the ID 12196 for showcasing this exporter's metrics.

You can also find the JSON file of the dashboard in this repository.

Note:

• This is just an example dashboard, feel free to use it as a base for your custom dashboard
• You need to adjust the color thresholds to suit your needs for the job status percentage singlestat panels
• Some panels might seem duplicated providing percentages and current values, these are just for showcasing the PromQL queries

Exported Metrics

RQ metrics:

Metric Name	Type	Labels	Description
rq_workers	Gauge	name, queues, state	RQ workers
rq_jobs	Gauge	queue, status	RQ jobs by queue and status
rq_workers_success_total	Counter	name, queues	Successful job count by worker
rq_workers_failed_total	Counter	name, queues	Failed job count by worker
rq_workers_working_time_total	Counter	name, queues	Total working time in seconds by worker

Request processing metrics:

Metric Name	Type	Description
rq_request_processing_seconds_count	Summary	Number of times the RQ data were collected
rq_request_processing_seconds_sum	Summary	Total sum of time spent collecting RQ data
rq_request_processing_seconds_created	Gauge	Time created at (time.time() return value)

Example:

# HELP rq_request_processing_seconds Time spent collecting RQ data
# TYPE rq_request_processing_seconds summary
rq_request_processing_seconds_count 1.0
rq_request_processing_seconds_sum 0.029244607000009637
# TYPE rq_request_processing_seconds_created gauge
rq_request_processing_seconds_created 1.5878023726039658e+09
# HELP rq_workers RQ workers
# TYPE rq_workers gauge
rq_workers{name="40d33ed9541644d79373765e661b7f38", queues="default", state="idle"} 1.0
rq_workers{name="fe9a433575e04685a53e4794b2eaeea9", queues="high,default,low", state="busy"} 1.0
# HELP rq_jobs RQ jobs by state
# TYPE rq_jobs gauge
rq_jobs{queue="default", status="queued"} 2.0
rq_jobs{queue="default", status="started"} 1.0
rq_jobs{queue="default", status="finished"} 5.0
rq_jobs{queue="default", status="failed"} 1.0
rq_jobs{queue="default", status="deferred"} 1.0
rq_jobs{queue="default", status="scheduled"} 2.0

Configuration

You can configure the exporter using command line arguments or environment variables:

CLI Argument	Env Variable	Default Value	Description
--host	RQ_EXPORTER_HOST	0.0.0.0	Serve the exporter on this host
-p, --port	RQ_EXPORTER_PORT	9726	Serve the exporter on this port
--redis-url	RQ_REDIS_URL	None	Redis URL in the form redis://:[password]@[host]:[port]/[db]
--redis-host	RQ_REDIS_HOST	localhost	Redis host name
--redis-port	RQ_REDIS_PORT	6379	Redis port number
--redis-db	RQ_REDIS_DB	0	Redis database number
--sentinel-host	RQ_SENTINEL_HOST	None	Redis Sentinel hosts separated by commas e.g sentinel1,sentinel2:26380
--sentinel-port	RQ_SENTINEL_PORT	26379	Redis Sentinel port, default port used when not set with the host
--sentinel-master	RQ_SENTINEL_MASTER	master	Redis Sentinel master name
--redis-pass	RQ_REDIS_PASS	None	Redis password
--redis-pass-file	RQ_REDIS_PASS_FILE	None	Redis password file path (e.g. Path of a mounted Docker secret)
--worker-class	RQ_WORKER_CLASS	rq.Worker	RQ worker class
--queue-class	RQ_QUEUE_CLASS	rq.Queue	RQ queue class
--log-level	RQ_EXPORTER_LOG_LEVEL	INFO	Logging level
--log-format	RQ_EXPORTER_LOG_FORMAT	[%(asctime)s] [%(name)s] [%(levelname)s]: %(message)s	Logging handler format string
--log-datefmt	RQ_EXPORTER_LOG_DATEFMT	%Y-%m-%d %H:%M:%S	Logging date/time format string

Notes:

• When Redis URL is set using --redis-url or RQ_REDIS_URL the other Redis options will be ignored
• When the Redis password is set using --redis-pass-file or RQ_REDIS_PASS_FILE, then --redis-pass and RQ_REDIS_PASS will be ignored
• The Sentinel port will default to the value of --sentinel-port if not set for each host with --sentinel-host or RQ_SENTINEL_HOST

Serving with Gunicorn

The WSGI application can be created using the rq_exporter.create_app() function:

$ gunicorn "rq_exporter:create_app()" -b 0.0.0.0:9726 --log-level info

Example Dockerfile to create a Docker image to serve the application with Gunicorn

FROM mdawar/rq-exporter:latest

USER root

RUN pip install --no-cache-dir gunicorn

USER exporter

ENTRYPOINT ["gunicorn", "rq_exporter:create_app()"]

CMD ["-b", "0.0.0.0:9726", "--threads", "2", "--log-level", "info", "--keep-alive", "3"]

Note about concurrency:

The exporter is going to work without any problems with multiple workers but you will get different values for these metrics:

• rq_request_processing_seconds_count
• rq_request_processing_seconds_sum
• rq_request_processing_seconds_created

This is fine if you don't care about these metrics, these are only for measuring the count and time processing the RQ data, so the other RQ metrics are not going to be affected.

But you can still use multiple threads with 1 worker process to handle multiple concurrent requests:

$ gunicorn "rq_exporter:create_app()" -b 0.0.0.0:9726 --threads 2

Building the Docker Image

$ # Build the docker image and tag it rq-exporter:latest
$ docker build -t rq-exporter .
$ # Or
$ make build

$ # M1 MacOS Build the docker image and tag it rq-exporter:latest
$ docker buildx build --platform linux/amd64 -t rq-exporter .

The image can also be built using docker compose:

$ docker compose build

Check the docker-compose.yml file for usage example.

Development

To start a full development environment with RQ workers, Prometheus and Grafana:

$ docker compose up
$ # If you want to start multiple workers use the --compatibility flag
$ # which will make docker compose read the `deploy` section and start multiple replicas
$ docker compose --compatibility up

You can access the services on these ports on your local machine:

• RQ exporter: 9726
• Redis: 6379
• RQ Dashboard: 9181
• Prometheus: 9090
• Grafana: 3000 (Login using admin:admin)

You can specify the services that you want to start by their name in the docker-compose.yml file:

$ # Example starting only the `rq_exporter` and `redis` services
$ docker compose up rq_exporter redis

To run more workers and enqueue more jobs you can scale the worker and enqueue services:

$ # Run 5 workers
$ docker compose up -d --scale worker=5
$ # Enqueue more jobs
$ # Scale the enqueue service and the workers
$ docker compose up -d --scale worker=5 --scale enqueue=2

To cleanup after development:

$ # Use -v to remove volumes
$ docker compose down -v

You can also start another rq-exporter instance that collects stats from a project using custom RQ Worker and Queue classes:

$ # Using -f to pass multiple docker-compose files
$ # docker-compose.custom.yml defines services using custom RQ classes
$ docker compose -f docker-compose.yml -f docker-compose.custom.yml up
$ # To cleanup you need to also pass the same files
$ docker compose -f docker-compose.yml -f docker-compose.custom.yml down

A new RQ exporter instance will be exposed on port 9727 on your local machine.

Note: If you don't have docker compose installed follow the installation instructions on the official website.

If you want to use the package manually:

$ # Clone the repository
$ git clone <REPO_URL>
$ # Change to the project directory
$ cd rq-exporter
$ # Create a new virtualenv
$ python -m venv /path/to/env
$ # Activate the environment
$ source /path/to/env/bin/activate
$ # Install the requirements
$ pip install -r requirements.txt
$ # Start the exporter on port 9726
$ python -m rq_exporter
$ # You can configure the exporter using command line arguments
$ python -m rq_exporter --port 8080

Running the Tests

$ make test
$ # Or
$ python -m unittest

Contributing

1. Fork the repository
2. Clone the forked repository git clone <URL>
3. Create a new feature branch git checkout -b <BRANCH_NAME>
4. Make changes and add tests if needed and commit your changes git commit -am "Your commit message"
5. Push the new branch to Github git push origin <BRANCH_NAME>
6. Create a pull request

Byrdification

Several customisations exists to make this run with minimal permissions. If you run this "Out the Box" then several write permissions are requested which we wanted to avoid. I tried monkeypatching but there is a chain of dependants so it didnt work.

We simply copy the edited/patched files in the Dockerfile to replace the real files in the rq library

1. rq.workers edited to avoid the need for srem permissions
2. rq.queue calls rq.registry which we edit to avoid the need for zremrangebyscore and match permissions
3. Permissions for the user needs to be: on ~* -@all +@read +select