[ntp:questions] server and peer lines correct?

Charles Swiger cswiger at mac.com
Mon Apr 7 18:05:03 UTC 2014


Hi--

On Apr 6, 2014, at 8:13 AM, Jason Rabel <jason at extremeoverclocking.com> wrote:
> Charles Swiger wrote:
>> Your comment means a discussion of best practices of configuring
>> these stratum-1s is also beyond the scope of answers to your post.
>> Odd, but have this as you would....
> 
> Charles, my S1 servers are a mixture of NTP appliances and other hardware that do not run the standard NTP distribution. They don't
> have the ability to configure peering and such. That's why I said it's beyond the scope of my post / question. ;)

That's fine.  It's unusual, but fine.  :-)

>> Don't bother using more than one iburst entry per server.
> 
> I think your wording is a little confusing. Are you saying only use iburst on one remote server and not all of them?
> 

> I was reading the NTP docs last night and came across a couple interesting notes. First it says not to use iburst with peer lines.
> Also it says the most stable behavior for symmetric active mode is to set maxpoll 6 for both peers. I'll probably give it a try and
> see how things turn out.

For a given S2 server, say NTP1, I'd recommend:

server S1-1 iburst prefer
server S1-2
server S1-3
server S1-4
server _some_outside_box_
peer NTP2
peer NTP3

Add the maxpoll 6 to the peer lines to incorporate the other suggestion.

NTP2 would have:

server S1-1 
server S1-2 iburst prefer
server S1-3
server S1-4
server _some_other_outside_box_
peer NTP1
peer NTP3

...etc.

>> Start everything up; wait a while; run?
> 
> What I meant was, is there a way to see / verify if the association with a remote machine is a peer-peer vs server-client? Other
> than looking in the configuration file? The regular ntpq -p output doesn't show any differentiation?
> 
>     remote           refid      st t when poll reach   delay   offset  jitter
> ==============================================================================
> +192.168.88.231  .GPS.            1 u  252  256  377    1.018   -0.212   0.246
> +192.168.88.232  .GPS.            1 u  239  256  377    1.018   -0.255   0.340
> *192.168.88.233  .GPS.            1 u   80  256  377    1.118   -0.234   0.250
> +192.168.88.234  .GPS.            1 u  196  256  377    0.960   -0.175   0.311
> +192.168.88.235  192.168.88.234   2 u   15   16  376    0.858   -0.179   0.070
> +192.168.88.60   192.168.88.233   2 u   39  256  377    0.407   -0.179   0.345

True; I don't think there's a specific billboard character to distinguish.
However, if you look at 192.168.88.233 - 235, each should show the other two hosts as reachable at stratum-2.

>> They'll be fine for several days.  However, if all of the S1s 
>> go down for an extended duration, and there are no other sources
>> of time available to the S2s, then the S2 servers will increase
>> their dispersion estimate, which will eventually cause downstream
>> clients to prefer other time sources.
> 
> There are no time sources listed in any configs that are outside of the local LAN. I supposed I could add some external NTP servers
> like you mention to my S2 group, but the thing that nags me is they are usually always off by a couple ms, I guess due to network
> delay and uncertainty and such.

Yes, but your local timeservers would likely be preferred due to lower delay and jitter, even without using a prefer statement.

If you do have a complete outage which lasts for many days, having an external timesource will eventually do better than undisciplined local clocks.  Depends on your connectivity, but selecting nearby external timesources generally gives me ones that seem to be synced to within 1-2 ms.

Regards,
-- 
-Chuck



More information about the questions mailing list