On Wed, Jun 19, 2019 at 11:55:51AM -0700, Watson Ladd wrote: > Dear fellow users, > > Since switching to chrony we've seen that the kernel's timex > sync_status flag occasionally becomes unsynced, while chronyd is > happily running and synchronizing the time acceptably. Unfortunately > we had an alert based on this back when were using ntpd. > > Can anyone shed light on how chronyd sets the sync_status flag when > running and why it might not update the kernel as frequently as ntpd?
chronyd sets the kernel status on each update of the clock, i.e. when the source marked with '*' has a good measurement that passed all the tests. The kernel automatically sets the "unsync" flag when the kernel's estimated maximum error (incrementing at 500 ppm) reaches 16 seconds. That takes less than 9 hours. If chronyd didn't have a good measurement in that time (e.g. due to a routing change in the network), the kernel unsync flag would be set. The difference between chronyd and ntpd is that chronyd can drop more than just 8 measurements in a row. ntpd has to update the clock occasionally, even if it doesn't have a good measurement to avoid the loop becoming unstable. I'll see if I can find a simple way to update the kernel status independently from the clock update. -- Miroslav Lichvar -- To unsubscribe email chrony-users-requ...@chrony.tuxfamily.org with "unsubscribe" in the subject. For help email chrony-users-requ...@chrony.tuxfamily.org with "help" in the subject. Trouble? Email listmas...@chrony.tuxfamily.org.
