David Woolley wrote:
Jaiprabhu wrote:
I have observed that the server reaches the reject state, while the
stratum stays the same (it was at 1), a long time after it moved into
the reject state. The server was moved to the reject state by making
it unreachable by adding it's IP in an iptables drop rule. The reject
state already means that the server is not a synchronization source.
What's the rule of thumb on the duration it would take for the stratum
shifting to 16?
It should shift when the last available sample is too old to really
correlate well with the true time. There is a definite drop dead when
the root dispersion reaches about a second. I thought that the stratum
collapses when somewhere between about 4 and 6 polls have been lost, but
it might not be until the estimated error (root dispersion) becomes too
large.
On my x86, running Ubuntu 9.04 and ntpd 4.2.4p4, it did not collapse
even after losing the 6th poll. I did not track the root dispersion but
I was tracking the time elapsed after all the configured hosts were made
unreachable for a while, and, it did not drop to stratum 16 till at
least 30 minutes after. I stopped keeping track after that.
_______________________________________________
questions mailing list
questions@lists.ntp.org
http://lists.ntp.org/listinfo/questions
