Thursday, November 7, 2019

NTP peer troubles

Ok gang, I don't usually need help, but this annoying situation has been gnawing at me off and on for a while now.

I'm confident one of you can nudge me in a good direction.

I ran this through TAC already, but let's just say communications issues interfered with our ability to find a complete solution.

I have four ASR1K WAN routers. Two per Data Center, times two data centers.

They all reach out to the interwebz for NTP.

Basic configuration looks like this:

ntp update-calendar clock calendar-valid clock timezone UTC 0 0 no clock summer-time ! ntp server A.A.A.A version 4 source loopback 123 ntp server B.B.B.B version 4 source loopback 123 ntp server C.C.C.C version 4 source loopback 123 ntp server D.D.D.D version 4 source loopback 123 ntp server E.E.E.E version 4 source loopback 123 ntp server F.F.F.F version 4 source loopback 123 ! ntp master 2 ! ntp panic update ! end 

Ok, so maybe six external sources is a tad excessive. Once upon a time (pun intended) we had a very bad experience with an NTP time warp. So, we're over-cautious, ok?

All of that works perfectly.
Each of my routers see all those sources and everything works perfectly at that level.
Each of my routers believes they are a Stratum-2 device, which is valid at least from an NTP topology perspective.

Here is the part that doesn't work the way I want it to work. (Probably because I'm doing something wrong)

What if I lost all of my internet connectivity?
I don't care that the total loss of internet connectivity is ridiculously far-fetched.
What if?

It feels to me like I should be able to use the ntp peer function to tell each router in the group about the other three, and they should be able to maintain an average time among themselves in the absence of a higher stratum source.

So I applied this configuration:

! ntp peer <Router-B> version 4 source loopback 123 ntp peer <Router-C> version 4 source loopback 123 ntp peer <Router-D> version 4 source loopback 123 ! end 

Some of my routers consider some of my other routers to be "insane", while some routers consider all 3 peers to be sane.

Am I mis-using the ntp peer feature? Should I just identify them as standard ntp servers instead of peers?

TAC was suggesting that the issue was with all 4 of my devices being configured as Stratum-2, but that didn't make sense to me.
In the NTP Hierarchy, they all have equally direct access to our upstream sources.

Thoughts? Suggestions? Kitty gifs?



No comments:

Post a Comment