[ntp:questions] Configuring a cluster: ntpd choosing local clock over server with lower stratum (and cycling between them)
david at djwhome.demon.co.uk
Thu Apr 5 07:11:36 UTC 2007
In article <1175732938.925035.105890 at o5g2000hsb.googlegroups.com>,
dromedaryl at yahoo.com wrote:
> .... And a few minutes later ...
Where a few minutes will be the very approximately 15 minutes needed to
verify a step change. If the local clock is the first one to become
reachable, it will lock onto that and then require the full broken clock
verification delay to change.
> 1) Why would ntpd choose to use its higher stratum system clock rather
> than a lower stratum server and why would ntpd cycle between them?
I'm not sure that it is worth analysing in detail in the light of the
following, but in the example you have only two clocks and their error
bands don't overlap, which is never a good thing. They also differ by
more than 128ms, so if ntpd locks onto one initially, it can take some
time to change to the other.
> 2) In the non-master node's ntp.conf, should I just remove the listing
> of its system clock as a server? Is it completely unnecessary as the
> non-master node is only a client?
Local clocks should never be configured on leaf nodes. They don't help
and they can cause confusion. They are over used on server configurations;
they inhibit the propagation of the indication that the time is unreliable.
> 3) In general, is the way I'm setting up the ntp.confs for the nodes
> on the cluster reasonable?
You should either remove the local clock from the server, or give it
enough upstream servers to outvote the local clock. (Local is treated in
a special way that might partially mitigate this, but I don't want to
trawl the code to be sure.)
More information about the questions