Hello!

On Fri, Mar 30, 2012 at 02:50:48PM +0200, Piotr Sikora wrote:

> Hi,
>
> >The major question seems to be "what should be done if we've not
> >yet sent full request, but already got response"? Right now line
> >is drawn at "if we've got response header, we stop sending
> >request". And if we move the line - we should clearly understand
> >where the new line is and why.
>
> While current approach seems acceptable for single request-response
> connections, I'm pretty sure that it can be source of corruption for
> keepalived connections (in case when
> backend doesn't read rest of the request after sending response or
> when backend wants to read rest of the request that wasn't
> completely sent).
>
> I see two possible solutions: either kill connection after such
> "incomplete" request is made or make sure to send whole request even
> if we receive response while doing so.
>
> Opinions?

I tend to think that connection should be closed in this case.
And it's actually likely to be closed by an upstream server
anyway.

Maxim Dounin

_______________________________________________
nginx-devel mailing list
nginx-devel@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx-devel