I have a network with two edge routers, one at each end, attached to independent gateways. I want to establish the edge routers as local time servers for the devices inside the network. The edge routers run the NTP package; the others make use of the included SNTP.
The edge router NTP clients work fine, setting the clock properly on the edge routers, but all the devices inside the network are receiving “stratum=0 KoD kiss of death” messages with the INIT explanation code, which is documented to mean that something isn’t initialized yet, but nothing really explains what it is, where it is, and what it needs.
Here is the setup at each edge router:
Screen Shot 2015-02-08 at 5.27.21 PM.jpg
I have tried every combination of multicast and manycast, and it makes no difference whatsoever.
Here is the setup at the non-edge routers:
Screen Shot 2015-02-08 at 5.30.00 PM.jpg
Here is what the non-edge routers receive:
Screen Shot 2015-02-08 at 5.32.54 PM.jpg
I have tried stopping and restarting each server, rebooting the units on both sides, and upgrading to 6.25 to get the “NTP fix,” and I can’t get past “server not synchronized.”
What’s annoying is that I have this same setup working fine on my main network… I remember having to work through a similar problem with NTP, but don’t really have any recollection of what I did that finally got it working.
You say “the edge router NTP clients work fine”, and yet the screenshot you provided is not in agreement with you: the clock may be set, but clearly the status of the client on the edge has not progressed from “started” when it should be “synchronized”. Also, the fact that your clock is either correct or close-to-correct may have nothing to do with the NTP client…if you are running this on x86, your machine most likely has its own battery-backed hardware clock that hasn’t drifted a whole lot since it was first set, and if you are running this on a RouterBoard, as recently as 6.16, RouterOS is now keeping time between reboots; from 6.16 changelog:
time – on routerboards, current time is saved in configuration on reboot
and on clock adjustment, and is used to set initial time after reboot
If the client on the edge router does not get to “synchronized”, the NTP server on the same device will answer its own clients in the way you have observed.
In the past, when I have faced similar problems, it has always turned out to be that I was trying to use a set of public NTP servers that are no longer up or valid. (I really wish the RouterOS NTP client didn’t try to resolve DNS hostnames to IP addresses at the time you configure the servers and would instead do a host lookup at least every bootup; both time.nist.gov and pool.ntp.org seem to rotate out servers on a fairly continuous basis, preventing me from configuring NTP at my edge on a “set-it-and-forget-it” basis.) However, it looks like the two IPs you are using belong to “phoenixpubliclibrary.org”, which I’m guessing is you or is at least under your control? So your problem isn’t likely to be that a public server you were previously using has been decommissioned. However, if your NTP client isn’t getting past the “started” phase on your edge router, then I would still point my blame at those two servers, and would be spending my time investigating the responses that your edge router is getting from them, not the responses that your non-edge routers are getting from the edge.
Ah, me! Here I was thinking that “started” meant “working,” as opposed to “stopped.”
And yes, now I remember, that this was exactly the problem that plagued me on the previous network – depending an advertised time server hosted by the Phoenix public library system (not these two particular servers, but another one that’s gone now). Never again! I switched to nearby NIST servers and everything came right back up.
I would have thought that if the edge client weren’t working, the other units would just get served whatever the edge router’s own clock contained, but it looks like if the edge client is enabled at all, the NTP package forces coordination between the client and server packages such that if the edge router ain’t happy, ain’t no one happy.
I switched to nearby NIST servers and everything came right back up.