prep for cutting a release, target 2019-01-13

[Achim Gratz]

Hal Murray via devel writes:
>> I'm certainly missing some context as I know neither what you tried nor what
>> test case you're talking about (presumably something in conjunction with
>> #437). 
>
> Sorry.  I was trying to replicate what I thought was your case.

No worries and thanks for looking into it.

> Server A was looking at several servers.  I turned one off.  The polling 
> interval ramped up to 1024.  I turned the server back on.  It took a while, 
> but the polling interval went back to 64.

What I'm looking at is seven NTP stratum-1 servers on the same LAN that all
monitor each other (they would be peers if peer more still existed) at
minpoll=maxpoll=4 (16s).

> Do you remember what your exact configuration was?

Recovery when one of the servers drops out (gets rebooted or
disconnected from the net or something like that) usually works.  What
doesn't work is that, once in a while (but more often when I start up
multiple servers in a short time) one of them decides it doesn't like
the questions of another.  I have not been able to actually see the
packet the target server sends, but the client status indicates that it
thinks it got a KOD packet from that server.  The server entry will hang
around and get polled at very long intervals (hpoll is mostly 10 when I
look at it).  The other interesting thing is that often only every
second poll then actually registers as successful in the client, making
the effective poll rate even longer.


Regards,
Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

SD adaptations for KORG EX-800 and Poly-800MkII V0.9:
http://Synth.Stromeko.net/Downloads.html#KorgSDada