nsayer@uop.uop.edu (Nick Sayer) (04/24/91)
We recently installed SunOS 4.1.1. After doing so, xntp's compliance number (reported by the "lo" command in xntpdc) has gone up by an order of magnitude. It now stays at about 0.001 or so. It used to sit at about 0.0002 or so. We also get "previous adjustment did not complete" messages in syslog. Before you mention it, yes I did run tickadj and put the resultant value in /vmunix. I also cleared dosynctodr. The hardware is a relatively heavily loaded Sun 3/160 with 12 MB of RAM and a pair of ALM-Is. At its worst, there are 10 or 12 users at a time, but there are lengthy periods when the system is completely idle. I would think those periods would have good compliance. We peer to a pair of stratum 1s on the internet, and we have a "stratum ceiling" driver for when the net goes down. :-/ -- Nick Sayer | Think of me as a recombinant | RIP: Mel Blanc mrapple@quack.sac.ca.us | Simpson: Homer's looks, Lisa's | 1908-1989 N6QQQ [44.2.1.17] | brains, Bart's manners, and | May he never 209-952-5347 (Telebit) | Maggie's appetite for TV. --Me | be silenced.
hakanson@CSE.OGI.EDU (Marion Hakanson) (04/25/91)
>We recently installed SunOS 4.1.1. After doing so, xntp's >compliance number (reported by the "lo" command in xntpdc) >has gone up by an order of magnitude. It now stays at about >0.001 or so. It used to sit at about 0.0002 or so. > >We also get "previous adjustment did not complete" messages >in syslog. You don't say what you were running before. I just started running xntpd under 4.1.1b, but on a SPARC IPC (lightly-loaded, so far). We do "tickadj -Asq" in rc.local right before starting xntpd. It shows compliance of -0.00002 or better, after a 24-hour settling in period (drift compensation is at about -0.2666, with the stock "tick" of 10000 -- I haven't tried 9998 as others have suggested for SPARCs). >The hardware is a relatively heavily loaded Sun 3/160 with 12 MB of >RAM and a pair of ALM-Is. At its worst, there are 10 or 12 users I have noticed that xntpd is more sensitive to a heavy load than the regular ntpd. And of course SunOS 4.x is worse (for clocks) than 3.5 was, but everyone knows that by now. But if all you did was go from 4.0.x or 4.1 to 4.1.1, I have no further explanation for the increased wobble -- all 4.x's have been pretty similar, NTP-wise, and 4.1.1 should be the best since 3.5. -- Marion Hakanson Domain: hakanson@cse.ogi.edu UUCP : {hp-pcd,tektronix}!ogicse!hakanson