Timeout without keep-alive

[imbolc]

Should not here be a check if keep-alive mode is active? https://github.com/KeepSafe/aiohttp/blob/master/aiohttp/server.py#L227

[fafhrd91]

this is different, it is slow request check. client can abuse connection and send header very slowly (kind of ddos)

[imbolc]

I mean if keep-ailve mode is off, shouldn't we count here only self._slow_request_timeout?

[fafhrd91]

i think it is just complicates code without benefits, also we run this code on first request as well when keep-alive state is not defined yet

[imbolc]

I mean the case in which tcp_keepalive disabled in app.make_handler. E.g.

app.make_handler(
    tcp_keepalive=Flase,
    slow_request_timeout=30
)

If I do not mention keepalive_timeout in this config it will be 75. So max(self._slow_request_timeout, self._keepalive_timeout) will be 75. But I expect 30.

[imbolc]

Maybe tcp_keepalive is redundant, why don't use keepalive_timeout=0 instead?

[fafhrd91]

actually tcp_keepalive is only viable option for aiohttp. keepalive_timeout is incredibly slow under load.

[asvetlov]

Handling keep-alives is required for supporting HTTP persistent connections.
But tcp_keepalive is very different, it works on TCP socket keepalive level (http://www.tldp.org/HOWTO/html_single/TCP-Keepalive-HOWTO/)

Thus we need both.

[imbolc]

Understood, thanks :)

[lock]

This thread has been automatically locked since there has not been
any recent activity after it was closed. Please open a new issue for
related bugs.

If you feel like there's important points made in this discussion,
please include those exceprts into that new issue.