[j-nsp] deleting ntp server from config, perhaps a bug?

[Chris Morrow]

On Thu, 27 Sep 2018 10:57:43 -0400,
Jared Mauch <jared at puck.nether.net> wrote:
> 
> 
> 
> > On Sep 27, 2018, at 10:51 AM, Saku Ytti <saku at ytti.fi> wrote:
> > 
> > On Thu, 27 Sep 2018 at 16:21, Chris Morrow <morrowc at ops-netman.net> wrote:
> > 
> >> there's (of course) the 'you should always deploy a full config' crew, but...
> > 
> > Reporting for duty.
>

I should be clear, the 'full config only' thing is good, but not
everyone is capable of getting there in the short-term. It also means
quite a bit of infra needs to exist, so not everyone is willing to
commit to it.

> The ability of routers to use DNS for service names is getting to be
> more of a soft-requirement => hard requirement these days.  I may
> want to configure a DNS name for my BMP/KFAFKA magic and have it
> fail over if we renumber the machine (for example).
>

intersting, what cadence would you expect for re-resolving named
resources though? hourly? at RR ttl expiry? other? Why is it not
acceptable to just run 'services' on a VIP, and not on the physical
machine's IP ? (ie: why would you ever renumber?)

> Cisco/Juniper require IPs for the NTP configuration, and if you type
> in something like a NTP pool name/label it will resolve it and store
> that in the config vs follow based on the pool measurement/accuracy
> over time.

see cadence question.