We were seeing occasionally 500 responses from nginx on our production servers. There is nothing in the error log correlated to the event. Upon turning on debugging log and TCP dump, we identified the occasionally 500 responses are caused by end user resetting the connection(especially end users who used IE). The error in the debug log showed as some thing like '2015/04/20 23:37:58 [info] 17423#0: *1188530641 writev() failed(104: Connection reset by peer)'.

Is this supposed to be a 499 error response instead of 500 error. When we see 500 errors, we usually think there is something wrong with the server. This is actually end user resetting the connection. Is there any plan to fix that in the future?

Thanks