[ntp:questions] Servers losing sync or not syncing at all.
mayer at ntp.isc.org
Sun Jan 1 04:24:13 UTC 2006
> I have two problems with my NTP server/client.
> The server & client are linux servers on a private lan.
> They are running NTP 4.2.0a
> The attached file showing the problem is ntp_nosync.txt
> 1) The client will not sync with the server at all. The server and client
> communicate perfectly as shown by the output of ntpq.
> The server is synced off a local GPS server and is rock steady but the
> just will not sync to it. I have included the output of tcpdump to show that
> server and client UDP packets and to my laymans eye they seem fine.
> The bd checksum can be ignored as tcpdump is getting the packet before the
> checksum is calculated thus showing it as invalid. However, the checksum is
> fine at
> the receiving end.
> The client syncs with the server for several minutes then for some reason it
> loses sync
> only to get it back again within a few minutes. This repeates forever. The
> server in this case
> is the GPS reference server. Now, if I use burst, the client stays synced
> forever, but given that
> the tcpdump show no communication problems, why should I need to use it.
> Anyway, any help would be greatly appreciated. Hopefully somebody who
> understands the NTP
> protocol and algorithms can determine from the tcpdumps what is really
> I am quite happy to run further tests if asked.
TCPdump shows a lot of UDP checksum errors. I'd check your hardware.
More information about the questions