Hacker News Reader

I run some pretty large NTP deployments and our NTP drift is never really over 1 second. Now the difference between NTP serverside drift and client server-side "skew" isn’t something I have looks at but for the most part I get a nagios alert on anything over 1 second of drift with 5 seconds being the most I have ever seen.

View on HN

Can you share more detail as to how you would alert on drift? What acts as the canonical source?

Not sure how you'd configure it, but you don't actually need a canonical source. You just compare it against all/many other servers and look at the difference. If there's any real skew you need to look at all of them that had a difference then but you don't need a canonical source anymore.

Reply on news.ycombinator.com