[ntp:questions] NTP architecture with virtual machines for stratum-2 ?

unruh unruh at invalid.ca
Mon Mar 4 20:20:46 UTC 2013

On 2013-03-04, janfrode at tanso.net <janfrode at tanso.net> wrote:
> I have access to two external stratum-1 ntp-servers (one GPS and one with both GPS and long wave radio), and one company internal stratum-1/GPS, plus I'm considering if we should have a second company internal stratum-1. The external ntp-servers are hosted at trusted sister organizations.
> So we have
> * ntp1 (GPS, sister organization 1)
> * ntp2 (GPS + radio, sister organization 2)
> * ntp3 (GPS, company internal)
> * ntp4 (not yet existing, do we need it? should we use something else than GPS?)
> and I'm wondering how best to distribute time for our servers. We use Red Hat Enterprise Virtualization. with kvm-clock for most servers, and this seems to have been keeping very good time so far. Will it be safe to use virtual machines for stratum-2, and connect all servers to these ? 

Virtualisation and time keeping should be kept as far apart as possible.
Time keeping should be done by the base OS, not by one of the virtual
ones. Otherwise you will get fights and bad times.

> Should our stratum-2 servers all be connected to ntp1-4, or is it better to have server1-ntp1, server2-ntp2, etc.. to make sure they don't all run off the same clock source? i.e. should we for server1 have ntp.conf with:

If source a gives good time, why should you worry about many machines
using it? The only purpose could be redundancy, and that you get by
having many servers, not by pointing them at different sources. 

>         server ntp1 prefer
>         server server2 
>         server server3
>         server server4
> or:
>         server ntp1 prefer
>         server ntp2 
>         server ntp3
>         server ntp4
> I can't point all our ntp-clients to the sister organizations stratum-1 servers, but I think our internal ntp1-appliances should handle the load of all our devices (a few thousand devices). Will it be best to point the ntp-clients to:

A few thousand should not be a problem, unless you get a company wide
electrical shutdown and all the machines come up at once. 

And what error budget do you have for the clocks. Do they all have to
have nanosecond accuracy, or one second accuracy?

>        server ntp3 prefer
>        server ntp4 prefer
>        server server1
>        server server2
>        server server3
>        server server4
> or just use the stratum2 servers?
>   -jf

More information about the questions mailing list