<div dir="ltr"><div class="gmail_extra"><br><div class="gmail_quote">On Mon, Jan 9, 2017 at 10:59 PM, Maxim Dounin <span dir="ltr"><<a href="mailto:mdounin@mdounin.ru" target="_blank">mdounin@mdounin.ru</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div id="gmail-:1zy" class="gmail-a3s gmail-aXjCH gmail-m15984d0fad8f1ccb">Typical kern.sched.quantum is about 100ms, so several<br>
CPU-intensive tasks can delay processing of the events enough to<br>
trigger a timeout if a context switch happens at a bad time.</div></blockquote></div><div class="gmail_extra"><br></div><div class="gmail_extra">Here what I see in truss' output:</div><br><div class="gmail_extra"><div class="gmail_extra">38.820523207 0.000006568 kevent(28,{ },0,{ 198,EVFILT_WRITE,EV_CLEAR,0x0,0x8218,0x821405071 },512,{ 6.215000000 }) = 1 (0x1)</div><div class="gmail_extra">39.783094188 0.000022875 kevent(28,{ },0,{ 52,EVFILT_READ,0x0,0x0,0x30b,0x81f800068 204,EVFILT_WRITE,EV_CLEAR,0x0,0x8218,0x821401588 51,EVFILT_READ,0x0,0x0,0xec,0x81f800000 68,EVFILT_READ,EV_CLEAR,0x0,0x8bf,0x81f816580 7,EVFILT_READ,EV_CLEAR,0x0,0x27f,0x81f813869 57,EVFILT_READ,EV_CLEAR,0x0,0x767,0x81f817bd8 203,EVFILT_READ,EV_CLEAR|EV_EOF,0x0,0x248,0x81f8030c1 181,EVFILT_READ,EV_CLEAR|EV_EOF,0x0,0x9b77,0x81f80ea68 178,EVFILT_READ,EV_CLEAR,0x0,0x39d,0x81f8010a9 198,EVFILT_READ,EV_CLEAR,0x0,0x3d3,0x81f805071 204,EVFILT_READ,EV_CLEAR,0x0,0x9da,0x81f801588 190,EVFILT_READ,EV_CLEAR,0x0,0x4ff,0x81f80fc48 154,EVFILT_READ,EV_CLEAR,0x0,0x88e,0x81f8130b1 151,EVFILT_READ,EV_CLEAR|EV_EOF,0x0,0xc1db,0x81f814290 157,EVFILT_READ,EV_CLEAR|EV_EOF,0x0,0xe841,0x81f80c029 195,EVFILT_READ,EV_CLEAR,0x0,0x952,0x81f8090a1 194,EVFILT_READ,EV_CLEAR,0x0,0x929,0x81f809ac8 201,EVFILT_READ,EV_CLEAR,0x0,0x4ef,0x81f80c980 174,EVFILT_READ,EV_CLEAR,0x0,0x51e,0x81f816518 77,EVFILT_READ,EV_CLEAR|EV_EOF,0x0,0x1168,0x81f811c61 },512,{ 5.253000000 }) = 20 (0x14)</div><div><br></div></div><div><br></div><div>1 second delay between two syscalls. Then nginx goes nuts processing all it missed during this second. I can not tell from this output how much time was spent in these syscalls. Can anyone?</div><div><br></div><div>What I don't like is timeout greater than 5 seconds. Doesn't it mean that system is allowed to block for timeout time to collect events?</div><div><br></div><div><br></div>-- <br><div class="gmail_signature"><div dir="ltr"><div>Руслан Закиров</div><div>Руководитель отдела разработки веб-сервисов</div><div><span>+7(916) 597-92-69</span>, <span>ruz @ <a href="http://www.sports.ru/" target="_blank"><img src="http://farm7.static.flickr.com/6235/6210250811_19a888dbba_o.jpg" width="43" height="14" style="vertical-align: bottom; margin-right: 0px;"></a></span></div></div></div>
</div></div>