On Thu, Dec 09, 2010 at 08:27:44AM -0800, Bill Unruh wrote: > On Thu, 9 Dec 2010, g...@tuxfamily.net wrote: > > > > >commit d1b820e7e323771d79a164ffa02cd9b835213bab > >Author: Miroslav Lichvar <mlich...@redhat.com> > >Date: Thu Dec 9 15:14:05 2010 +0100 > > > > Shorten initial delay and sampling separation to 0.2 seconds > > Is this going to run into problems with ntpd's Kill signal, thinking that your > machine is flooding the server (or is the current behaviour of chrony already > in danger of that anyway)?
It should be ok. The 0.2s separation is to limit the rate of our requests to all configured servers, so we don't clog our link. For individual servers the spacing still follows the polling interval in normal mode and 2s interval in burst mode. The previous commit fixed a bug where chrony on start sent most of the requests at the same time, because the scheduling separation ignored timeouts that were scheduled for the exact same time. BTW, ntpd seems to separate the request only on start, by 1 second (which is the resolution of its timer), and then tries to minimize the chance of two requests meeting in the same 1s tick of the timer by adding some randomness to the interval. Chrony now adds some randomness too (2%), but that's mainly to avoid a group of clients to synchronize their requests' times. Probably very unlikely to happen even without the extra randomness as the polling interval includes the round-trip time of the previous request which is random. -- Miroslav Lichvar --- To unsubscribe email chrony-dev-requ...@chrony.tuxfamily.org with "unsubscribe" in the subject. For help email chrony-dev-requ...@chrony.tuxfamily.org with "help" in the subject. Trouble? Email listmas...@chrony.tuxfamily.org.