Home / server / Questions / 1056304
Accepted
molik
Asked: 2021-03-09 06:39:25 +0800 CST

Can a NTP Server escape the falseticker status?

  • 772

I have 2 ntp server syncing to their local clock, the first one is a stratum 1 and the second is a stratum 7. The client is polling both of them and don't have a local clock as backup.

My ntp results are the following (view from the client) :

     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
*192.168.1.5     .GPS.            1 l    1    32  377    1.492   -0.107   0.249
+192.168.1.6     LOCL            8 l    1    32  377    1.369   183.293  0.249


     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
x192.168.1.5     .GPS.            1 l    1    32  377    1.383   -0.135   0.120
x192.168.1.6     LOCL            8 l    1    32  377    1.3677  183.231  0.122

I know in my case that the difference between the offsets are very big. I am just testing things, theses servers won't be used.

If a server is declared as a falseticker by the client, can the status be reconsidered ? Or is the server sentenced to stay falseticker ?

I know after a restart of the ntp service on the client, the server status will be refreshed. But can it happen without restarting ntp on the client ?

ntp

1 Answers

  1. Best Answer

    NTP falsetickers are statistical outliers in the last few time stamps. The status can change when the algorithms evaluate new data. You don't have the 4+ NTP servers required to reliably detect falsetickers. That may not be a problem if you don't have demanding requirements, could just use one decent performing NTP server.

    Pay attention to the magnitude of the statistics. An offset of 180 ms is enormous when error from delay and jitter is single digit ms.

    192.168.1.6 with refid LOCL is a little unusual. Undisciplined local clock pointing to self should not be used. But usually the LOCL driver is invoked with config lines matching server 127.127.1. Is that a remote NTP server? If remote, and you are overriding the refid string in ntpd.conf, don't.

    On some lightly loaded host, configure ntpd or chrony to add additional references from internet sources. As a test, although a diverse source list is a great start if you wanted to build stratum 2 NTP servers.

    server 192.168.1.5 iburst
server 192.168.1.6 iburst
pool 2.pool.ntp.org iburst
server time.cloudflare.com iburst
server time.facebook.com iburst

    Anything flagged as a falseticker in this local plus internet configuration likely is a real outlier.

    Your NTP servers on the LAN should be very close to each other, better than 10 ms. They should reference each other. For example, configure the .6 appliance to reference .5 as a NTP source, and vice versa.

    • 2