RuntimeError with aiobotocore + Tornado about aiobotocore

[thehesiod]

from what I see this is an issue with your usecase/tornado. aiobotcore and its dependencies require the use of an Event loop that's compatible with BaseEventLoop. The BaseEventLoop requires support of Tasks and a task requires a current_task and it doesn't seem like tornado sets/provides this. btw even passing timeout=None would cause the error because it does the current_task check before the None timeout check in that timeout class.

[thehesiod]

btw, can you try with the latest aiohttp? The codepath has changed significantly in regard to timeouts

[andymo]

Regarding Tornado's use of current_task, it was brought up in the two issues I linked previously, as well as most other issues opened regarding tornado + aio libraries. I figured since a fix did make its way into aiohttp, aiobotocore could support it. Otherwise I understand, it's not really aiobotocore's responsibility :)

I am on the latest pypi version of aiohttp (1.3.3), which should have all commits regarding timeouts that were pushed to fix aio-libs/aiohttp#1180, since I used this example and it was working. I see in the changelog of 1.3.4 that there is a revert of some timeout handling in client request, so I'll pull that and see if that helps, thanks!

[adamrothman]

@andymo The simplest fix is to just wrap your calls in a Task, like so:

async def get(self):
    session = aiobotocore.get_session()
    client = session.create_client('s3')

    task = asyncio.ensure_future(client.list_buckets())
    response = await task

    self.write(response)

[thehesiod]

btw, I'm disabling the use of aiohttp's read_timeout (#248) as it is implemented incorrectly, this may help your cause. If getting rid of conn timeout helps as well we can look into that. I'm really unhappy with their timeout impl :(

[jettify]

@thehesiod do you think we can close this?

[thehesiod]

I believe so