[ntp:questions] Is there any non-obvious reason why NTP won't increase the polling rate on a new system?

Pete Stephenson pete+usenet at heypete.com
Tue Jul 24 15:32:56 UTC 2007

In most of the local stratum 2-3 servers I've setup for the first time, 
they spend an hour or two querying the remote servers at 64s. Within a 
few hours, they increase the polling time up to 1024s and there they 
remain barring any major problems.

Yesterday I configured a new stratum 2 system[1]. With a basic ntp.conf 
file[2] it fetches time and syncs the clock to the proper time within a 
few milliseconds. However, even after running for 13+ hours, it refuses 
to increase the poll interval higher than 64s.

The frequency is reported as -17.264 in the ntp.drift file.

The network is asymmetrical (cable) but stable and relatively 
low-latency (delays range from about 33ms-53ms for the remote server). 
Jitter is between 1.5-3.8. Reach values for each server are 377. Offsets 
range between 0.4-6.1ms, which isn't bad. The network is completely 
unused except for outgoing NTP queries and occasional checking of email. 
There've been no periods of unavailability, high latency, or other 
network issues.

The system's been running for 13 hours with no significant load other 
than that provided by ntp (load averages are 0.13, 0.10, 0.06). The 
computer is within an air-conditioned room with a relatively stable 
temperature of ~70F.

I've checked everything(?) that's would obviously cause problems 
(network problems, temperature swings, etc.). Everything looks to be 
proper and conducive to good working order. Is there anything that I 
might be missing that would result in ntpd not increasing the polling 
interval? I haven't restarted the service or the computer, as I figured 
that ntpd would work itself out once time went on.

Something seems borked, and I don't want to pester the upstream time 
servers[3] with queries every minute.

Any suggestions?

[1] Ubuntu Linux 7.04. Running ntpd v4.2.2p4, the version installed from 
the Ubuntu packages server.
driftfile /var/lib/ntp/ntp.drift
restrict default kod notrap nomodify nopeer
restrict nomodify
[various lines related to statistics]
server nist1-sj.witime.net iburst prefer
server pubts0-sj.witime.net iburst
server clock.sjc.he.net iburst
server time.sonic.net iburst
server tick.qwest.net iburst
server ntp2.mainecoon.com iburst
[3] Does anyone have any means of reaching the time server admins at 
witime? Their pubts1-sj.witime.net server is seriously borked, and the 
ntp.conf files of their other public stratum 2 servers are misconfigured 
-- instead of peering with each other, they're peering with non-existent 
systems on Time Warner Telecom's network due to someone mistyping the IP 
addresses. I wrote to the contact address on their website (info@), but 
have yet to receive any response. Fortunately, their stratum 1 servers 
they're running for NIST are working all right.

Pete Stephenson

More information about the questions mailing list