Prometheus Async

26.1.0 · active · verified Thu Apr 16

prometheus-async provides asynchronous helpers for the `prometheus_client` library, making it easier to expose metrics from `asyncio` and `Twisted` applications. Maintained by Hynek Schlawack, the library is currently at version 26.1.0 and receives regular updates, often focusing on type hinting and internal improvements, with occasional breaking changes for major Python versions.

Common errors

```
AttributeError: module 'prometheus_async.aio' has no attribute 'start_http_server'
```
cause Incorrect import path for `start_http_server` in the `aio` module.

fix Change the import to `from prometheus_async.aio.web import start_http_server`.
```
TypeError: start_http_server() got an unexpected keyword argument 'loop'
```
cause Attempting to pass the `loop` argument to `start_http_server()` on a version where it has been removed.

fix Remove the `loop=...` argument from your call to `start_http_server()`. The library handles the event loop automatically.
```
ModuleNotFoundError: No module named 'prometheus_async.aio'
```
cause `prometheus-async` library is not installed or there's a problem with your Python environment's path.

fix Run `pip install prometheus-async` to install the library. If already installed, check your virtual environment activation or `PYTHONPATH`.

Warnings

breaking Dropped support for Python 2.7, 3.5, and 3.6. `prometheus-async` now requires Python 3.9 or newer.
Fix: Upgrade your Python environment to 3.9 or a later version. For older Python versions, use `prometheus-async<22.1.0` (not recommended for new projects).
breaking The `loop` argument has been removed from `prometheus_async.aio.web.start_http_server()`.
Fix: Remove the `loop` argument when calling `start_http_server()`. Modern `asyncio` manages the event loop automatically, making explicit `loop` passing generally unnecessary and discouraged.
gotcha Ensure you import from the correct asynchronous backend (`aio` for asyncio/aiohttp or `twisted` for Twisted). Mixing them or using the wrong one will lead to `AttributeError` or unexpected behavior.
Fix: Verify that your imports (e.g., `prometheus_async.aio.web.start_http_server`) match the asynchronous framework you are using in your application.

Install

pip install prometheus-async Install latest version

Imports

start_http_server
wrong:
```
from prometheus_async.aio import start_http_server
```
correct:
```
from prometheus_async.aio.web import start_http_server
```
The `start_http_server` function for asyncio is located in the `web` submodule, not directly under `aio`.
time
```
from prometheus_async.aio import time
```
Common decorator for timing async functions.
track_inprogress
```
from prometheus_async.aio import track_inprogress
```
Common decorator for tracking in-progress async operations.
monitor
```
from prometheus_async.aio import monitor
```
Generic decorator for monitoring async functions with custom metrics.

Quickstart

This example demonstrates how to start a Prometheus HTTP server using `prometheus_async.aio.web.start_http_server` and how to apply the `prometheus_async.aio.time` decorator to an asynchronous function. The metrics server runs in the background, making metrics available at `/metrics` on the specified port.

import asyncio
from prometheus_client import Gauge
from prometheus_async.aio.web import start_http_server
from prometheus_async.aio import time

# Define a metric
MY_GAUGE = Gauge('my_async_gauge', 'Description of my async gauge')

@time(MY_GAUGE)
async def some_async_function():
    """An example async function that does some work."""
    print("Running some_async_function...")
    MY_GAUGE.set(10)
    await asyncio.sleep(0.5)
    MY_GAUGE.set(20)
    print("some_async_function finished.")

async def main():
    # Start the Prometheus HTTP server in the background
    # It will expose metrics on http://localhost:8000/metrics by default
    server = await start_http_server(port=8000)
    print("Prometheus metrics server started on port 8000")

    # Run our async function
    await some_async_function()

    print("Application finished. Metrics server still running. Press Ctrl+C to exit.")
    # Keep the event loop running to allow the metrics server to serve requests
    await asyncio.Future() # This will run forever until cancelled

if __name__ == "__main__":
    try:
        asyncio.run(main())
    except KeyboardInterrupt:
        print("\nExiting.")

view raw JSON →