Score:1

NTP Client not Syncing to Local Server

ro flag

We have an ntp server run on a Windows machine which we want the debian device to sync from. I have tried numerous other suggestions in similar questions, none of which have had any results. This includes

  • no ufw present
  • only the single line "server 10.60.58.102" in ntp.conf
  • "tinker panic 0" at start of ntp.conf
  • "burst" "iburst" "prefer" "flag1 1" flags for the server/fudge lines
ntpq -p
     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
 10.66.58.102    .LOCL.           1 u   36   64    1    0.715  208264.   0.002
 0.debian.pool.n .POOL.          16 p    -   64    0    0.000    0.000   0.002
*combine-window. 110.142.180.39   2 u   31   64    1  141.888    9.062  12.209
 speedtest.cribb 139.99.222.72    3 u   29   64    1  147.905    9.737  11.597

As can be seen from above, the offset is approximately 208 seconds to the desired server which it refuses to sync from. It never has the asterixis next to this server for any of the ntp.conf configurations that I have tried. Although it does appear to sometimes lower the offset over time by about ~5-10ms occasionally (once every minute or so).

Any help getting it to sync with the desired LAN NTP server appreciated.

Marcus Müller avatar
pt flag
this sounds like it's trying to controlledly catch up, avoiding a sudden jump in time
vbscript avatar
ro flag
Possibly, but I tried with the "tinker panic 0", "burst" and "flag1 1" which from reading other questions should help with this issue of large time discrepancies.
Score:1
cn flag

Your NTP is functional. LAN NTP server was discarded because it does not agree with internet sources.

First character of peers from ntpd -p is the tally code. * means system peer, the best one according to the algorithms. reach 1 implies ntpd was just restarted, need to wait a few minutes for the first few packets, and for all the pool servers to be resolved.

The two internet sources agree: about 9 ms offset from this host, plus or minus less than 1 ms. And good performing, low jitter for hosts 150 ms away. In contrast, 10.66.58.102 is 208,000 ms offset. That enormous difference cannot intersect the small error bars of the other two.

Fix the Windows box to use a well performing reference clock. A public NTP service can work for this, such as the NTP Pool project, same as the Linux host.

vbscript avatar
ro flag
Thanks, but even without those other servers being specified in the config file, the desired ntp server it not used and discarded, despite the "tinker panic 0" and "flag1 1" options tried at the top of the file and after the fudge command respectively. I want it to follow the Windows NTP server regardless of how right/wrong that time is.
John Mahowald avatar
cn flag
Your peers output has other NTP associations than 10.66.58.102, which is not what you described. Delete server and pool lines you do not use. And restart ntpd.
vbscript avatar
ro flag
You are correct, the text I had was not what I described in my comment, but instead it was just one of numerous outputs from the many tests I have run. I have removed all unneeded server and pool lines, and over the course of 5 hours it went from 20731 to 205946 offset (~1 second). It only had the two lines: "tinker panic 1" and "server 10.66.58.102 prefer burst". I was hoping it would change faster than 1 second in 5 hours, (not sure if tinker panic 0 or 1 is what I want but neither seem to force instant update).
John Mahowald avatar
cn flag
`ntpd -g` command line option allows an unlimited step, but only once. Restart with the -g option to step immediately.
mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.