> Thanks. That is correct - I was looking for Sun specific implementation (as > I am new to Solaris) synchronizing with a reference clock and a common queue > of events that can be accessed by all peer servers in the network. For > instance, does it use any specific algorithm such as Lamport's for clock > synchronization? >
I do not know if I can help you with the algorithms. I think that the source is open and you can go look at it for that sort of detail. If you want NTP to work in your network then you need some stratum 1 or stratum 2 time sources or a locally attached radio clock. Unless you happen to know all the details of setting up and then maintaining your own cesium based atomic clock then I suggest that you locate some time sources near you. Please check with your local universities as they generally have an NTP server. Also, your country should have a stratum 1 level time source also. So locate three sources and get their ip addresses or host names. Then create an ntp.conf that looks like so : $ cat /etc/inet/ntp.conf server time.server.number.one server time.server.number.two server time.server.number.three driftfile /etc/inet/ntp.drift statsdir /var/ntp/ntpstats/ filegen peerstats file peerstats type day enable filegen loopstats file loopstats type day enable filegen clockstats file clockstats type day enable logconfig =syncevents +peerevents +sysevents +allclock The start the NTP service with svcadm. $ svcs -av | grep ntp online - Mar_02 36 svc:/network/ntp:default After a little time you will begin to get NTP data flowing in and your server will self adjust to become ever more accurate and then keep itself there. $ /usr/sbin/ntpq -c rl status=0654 leap_none, sync_ntp, 5 events, event_peer/strat_chg system="SunOS", leap=00, stratum=3, rootdelay=25.92, rootdispersion=102.08, peer=10660, refid=time1.chu.nrc.ca, reftime=c996e92d.4826b000 Mon, Mar 5 2007 14:08:29.281, poll=10, clock=c996eb95.1f9eb000 Mon, Mar 5 2007 14:18:45.123, phase=-12.839, freq=-625.76, error=41.99 $ /usr/sbin/ntpq -pn remote refid st t when poll reach delay offset disp ============================================================================== *209.87.233.53 209.87.233.51 2 u 620 1024 377 24.43 -26.910 7.37 +128.100.100.128 128.100.96.226 2 u 530 1024 357 72.72 -2.350 12.15 +128.100.56.135 128.100.96.226 2 u 922 1024 377 65.95 -11.023 13.52 $ Then, other machines in your network can get time data from your first server. Just create an ntp.conf thus : $ cat /etc/inet/ntp.conf server my.time.server.host driftfile /etc/inet/ntp.drift statsdir /var/ntp/ntpstats/ filegen peerstats file peerstats type day enable filegen loopstats file loopstats type day enable filegen clockstats file clockstats type day enable logconfig =syncevents +peerevents +sysevents +allclock then start ntp services. $ /usr/sbin/ntpq -c rl status=0644 leap_none, sync_ntp, 4 events, event_peer/strat_chg system="SunOS", leap=00, stratum=4, rootdelay=26.40, rootdispersion=120.70, peer=3540, refid=my.time.server.host, reftime=c996ec1e.77775000 Mon, Mar 5 2007 14:21:02.466, poll=6, clock=c996ec28.23157000 Mon, Mar 5 2007 14:21:12.137, phase=2.623, freq=10100.77, error=0.98 You can do this with a stack of servers. If you feel so inclined you can drag a fair amount of data out of that stats information collected. see http://www.blastwave.org/ntpstats/index.html Dennis _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org