Circular dependency between local_unbound and ntpd?

Paul Mather paul at
Tue Jul 14 13:48:43 UTC 2015

I believe I ran afoul of a circular dependency between local_unbound and ntpd on my 10.2-PRERELEASE system.  I use a stock /etc/ntp.conf and use ntpd_sync_on_start="YES".

Last night, a BIOS settings reset cause my CMOS clock to go WAY out of synch for the first time.  No problem, I thought: NTP will correct it at boot.


When my system booted, the time was not corrected.  Also, DNS resolution was not working.  I figured out it was because local_unbound relies on an accurately set clock, but the clock could not be set accurately because my stock ntp.conf requires working DNS resolution to reach the NTP servers.

That sounds like a potential circular dependency to me.

My workaround at the time was to look up on another system; stop ntpd; then do a ntpdate using the IP addresses to set the clock. Once the clock was set accurately, things were all hunky dory.

Does anyone have any suggestion for an automatic way around this?  I guess one way would be to put the IP address of an NTP server into my ntp.conf file, so at least one would be reachable without needing a working DNS?

My main concern is for those systems like my Raspberry Pi and Beaglebone Black that don't have a battery-backed clock.  I currently don't use local_unbound on those, but it seems like I'd encounter this problem routinely if I did.



More information about the freebsd-stable mailing list