Changing monitoring system to do multiple queries

Testing the new monitoring daemon I noticed that unexpectedly many servers seem to drop return “RATE” (rate limiting / KoD) response when the client does more than 2 (or a few more) queries.

I think it’s reasonable for a client to do multiple queries; either as “iburst” in ntp.conf or as an sntp client sampling to minimize network jitter.

With that in mind I plan on increasing the number of queries done to each server from the monitoring system(s) in the not too distant future. I’ll attempt to note the “rate” responses in the monitoring logs to make the results from this change easier to debug.


FWIW, i turn on rate limiting in ntpd while using the default limit settings, or slightly more generous ones. (It defaults to off, but if you turn it on, there are default numbers.) iburst or (recent versions of) ntpdate should be fine, other very fast clients may not be.

(Warning: Older versions of ntpd used different units for one of the discard settings.)

I think what I will do is setup two monitors in the new pool that run on the same system but have different “number of samples” settings.

For much servers more samples should give better results (assuming the monitoring daemon is good about picking the best sample…).

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.