Potential starvation with connection pool

Open
#1,247 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
35/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Stale
Tech stack
postgresql, python
Domain
databases

Research direction

Start by running the provided Python reproduction with asyncio.TaskGroup, 100 workers, and asyncpg.pool.create_pool(max_size=10). Inspect the behavior of pool.acquire(timeout=10) under contention and determine whether the observed TimeoutError reflects starvation or expected scheduling; done means documenting the fairness guarantee or identifying a reproducible pool defect.

Written by the indexing model from the issue text.

Description

I have the following piece of code where I initialize 100 workers and a connection pool of size 10. Each worker loops and trys to acquire a connection and hold it for 200ms.

import asyncio

import asyncpg


async def worker(pool: asyncpg.Pool):
    while True:
        async with pool.acquire(timeout=10):
            # hold the connection for 200ms
            await asyncio.sleep(0.2)


async def main():
    async with asyncpg.pool.create_pool(max_size=10) as pool:
        async with asyncio.TaskGroup() as tg:
            for _ in range(100):
                tg.create_task(worker(pool))


asyncio.run(main())

When running on my laptop, this code reliably crashes with a TimeoutError after a few minutes, meaning that at least one worker failed to acquire a connection after 10 seconds. Given that there're 10 workers per connection and each worker holds the connection for 200ms, I would expect workers to be able to acquire a connection every 2 seconds.

This looks like starvation to me. Are there fairness guarantees regarding connection pools? Is there anything I could do other than just increasing the max pool size?

Dominant language
Python
Stars
8.1k
Forks
469
Avg merge
2d 20h
Merged PRs (30d)
9

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from MagicStack/asyncpg

All issues in MagicStack/asyncpg

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.