On 26.2.2008, at 12.11, William Herrin wrote:
If the NTP client regularly communicated with the server, this wouldn't be a problem. It would see high jitter on the initial attempt which would settle right down. But NTP is a well behaved protocol. When the server is responding well and it's in sync, NTP backs off to where it sends only a single query every thousand seconds or so. That's more than enough time for the ITR to expire and remove the cached ETR entry so that on the next attempt NTP sees: jitter.
That 2^10 seconds is just default maximum; protocol allows for configurations to back off up to 2^17 seconds (day and half).. That type of configurations would break in any system that isn't pure push, due to potential latency being either the packet roundtrip time, or the lookup time + packet transit time.
Of course, NTP would recover from that (I believe; I haven't verified by reading the code.. err, 'protocol' draft) by marking the peer insane, winding up using someone else for awhile, and then possibly restarting using peer with the backoff timer starting from scratch. Still, it wouldn't be pretty.
Cheers, -Markus -- to unsubscribe send a message to rrg-request@psg.com with the word 'unsubscribe' in a single line as the message text body. archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg