Beware the observation bias. Maybe try going back to v1.6.2 (or v1.6.3 for
the latest 1.6)?

These messages indicates the remote side resets the connection, not nginx,
thus your PHP-FPM is responsible for that.
Why? Ask it :o)
---
*B. R.*

On Fri, Dec 4, 2015 at 12:58 AM, silentmiles <nginx-forum@nginx.us> wrote:

> I have also started to see these errors in my log -- readv() failed (104:
> Connection reset by peer). I recently upgraded from nginx 1.6.2 to 1.8, and
> I think they started from this point.
>
> This is with a Linux/PHP-FPM setup. I've found a few mentions of the above
> error with this setup, but they're all related to back-end timeouts. None
> of
> my requests to the back-end run for more than a second or two, so I don't
> think this is a timeout issue for me.
>
> Has anyone else seen this issue occur recently, or can provide any
> pointers?
>
> Posted at Nginx Forum:
> https://forum.nginx.org/read.php?2,262035,263252#msg-263252
>
> _______________________________________________
> nginx mailing list
> nginx@nginx.org
> http://mailman.nginx.org/mailman/listinfo/nginx
>
_______________________________________________
nginx mailing list
nginx@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx