Greer Reichow wrote: > All, > I've come across aan interesting problem I'm hoping someone has seen > > I have the following in a development environment: > 1)T2000 - 8 core, 32GB RAM > 2)T5120 - 8 core, 32GB RAM > 3)Symmetricom S250 GPS NTP Server > > Both the T2k & 5120 are built off the same jumpstart images (solaris 08/07) > with ~ 4 ldoms each. Each LDOM is configured with 4CPUs and 2 GB RAM. We are > running ntp client version 4.2.4 downloaded from sunfreeware.com. LDOM level > is 1.0.1 currently > > Each LDOM on both machines is running the same ntp.conf file: > server <ip> minpoll 4 maxpoll 4 prefer > driftfile /var/ntp/ntp.drift > > While I know this wouldn't be ideal in a large deployment, the fact that I > have a private DNS stratum 1 server and few clients allows me to set the time > polling fairly tight. > > Here is the interesting part: > Results of ntpq -c peers > > 5120: > delay offset disp > .061 37.037 0.12 > > T2000: > delay offset disp > .062 .04 0.14 > > Why am I getting such disparate results in offset? The project I'm > developing this for has very strict timing requirements that the T2000 seems > to meet (< 10ms accuracy) that the newer 5120 doesn't meet. Has anyone seen > something similar? Right now this is pointing at a hardware issue, but I > want to eliminate the ldom code as a problem
There are clock drifts on the systems that are greater than what NTP can handle. Came across this myself just last week. There are /etc/system settings described in Change Request 6630235 based on CPU clock frequency. I am looking for a public document that describes these. Steffen > Thanks, > Greer > -- > This message was posted from opensolaris.org > _______________________________________________ > ldoms-discuss mailing list > ldoms-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss
