[ntp:questions] orphan mode and layered redundancy
florin at andrei.myip.org
Thu May 26 22:04:30 UTC 2011
A few local NTP servers, feeding off of the US pool. Multiple clients,
grouped in clusters; it is important that the cluster members stay
synced to each other. I could use ntp-4.2.6p3, but more recent than that
would be problematic.
If the Internet feed goes offline, I want the local NTP servers to stay
in sync with each other.
But also, if the local NTP servers are shut down (unlikely in
production, but will happen once in a while in the testing environment),
I want the members in each cluster to stay in sync with each other for
the duration of the planned NTP outage.
This is my proposed configuration, please criticize it:
(I only represented one cluster with nodes C and D, but in reality there
are several independent clusters pointing to the local NTP servers.)
It is unclear to me whether "tos orphan N" needs to have different
values between NTP servers and cluster members. I assumed it needs to be
higher for cluster members.
Do I have to add "tos cohort 1" anywhere? Or would the "peer" commands
take care of everything?
More information about the questions