On 8/8/2016 6:14 PM, Jacob Champion wrote:
On 07/25/2016 11:13 AM, Arthur Ramsey wrote:I think I will try the following settings first, but failing that I'll give the x86 build a try. AcceptFilter https noneAny follow-up on this? I've been digging into the AcceptEx() implementation, since it looks like there have been intermittent problems with it for a while now. If you have a reproduction case, I'd be happy to run with it. --Jacob --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscribe@xxxxxxxxxxxxxxxx For additional commands, e-mail: users-help@xxxxxxxxxxxxxxxx
We've also seen this in our 64-bit Apache 2.4.16 build on Windows. Specifically, accessing/refreshing a particular page in Chrome would lead to a 10-20 second hang for all clients, Firefox would maybe lead to a 2-5 second hang sometimes, but IE had no effect. The page had a mixture of static files (mainly JavaScript) and dynamic content loaded via Ajax. I was able to determine from the access log that the actual request processing took a normal amount of time, so the extra time must have occurred before the request start time was recorded (i.e. during accept/reading the request line).
Though in our case, we only needed to use AcceptFilter http connect AcceptFilter https connectrather than turning it off completely using "none". Setting it to connect allows the server to recycle sockets.
I'll see if I can't look back into it and try to find a reproduction case again that I can narrow down.
-- Paul Spangler LabVIEW R&D National Instruments --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscribe@xxxxxxxxxxxxxxxx For additional commands, e-mail: users-help@xxxxxxxxxxxxxxxx
|