Bug #7108
closed
ntp does not keep time on virtualized pfsense
0%
Description
I run pfsense virtualized under FreeBSD bhyve. I've read all of the normal advice and have implemented the usual workarounds but can't seem to get ntpd from base to stay synched more than a few hours.
I've implemented a workaround of calling ntpdate -u every 15 minutes or so which seems to help ntpd stay in synch. I realize pfsense switched away from openntpd some time back but I have to say it does a vastly superior job of keeping time synched in a VM. Is this something that could be added as an alternative, especially since it's significantly less buggy (e.g., more secure) than ntpd from base?
/boot/loader.conf.local
kern.timecounter.hardware="ACPI-fast" kern.hz="100"
Updated by John Silva over 7 years ago
Relevant logs:
Jan 7 23:18:09 pfsense ntpd[86455]: 192.168.1.10 8011 81 mobilize assoc 23512 Jan 7 23:18:10 pfsense ntpd[86455]: 192.168.1.10 8024 84 reachable Jan 7 23:18:10 pfsense ntpd[86455]: 192.168.1.10 903a 8a sys_peer Jan 7 23:18:10 pfsense ntpd[86455]: 0.0.0.0 c614 04 freq_mode Jan 7 23:23:46 pfsense ntpd[86455]: 0.0.0.0 0612 02 freq_set kernel -625.128 PPM Jan 7 23:23:46 pfsense ntpd[86455]: 0.0.0.0 061c 0c clock_step -0.210043 s Jan 7 23:23:46 pfsense ntpd[86455]: 0.0.0.0 0615 05 clock_sync Jan 7 23:23:46 pfsense ntpd[86455]: frequency error -625 PPM exceeds tolerance 500 PPM Jan 7 23:23:49 pfsense ntpd[86455]: 0.0.0.0 c618 08 no_sys_peer Jan 7 23:23:49 pfsense ntpd[86455]: 192.168.1.10 8044 84 reachable Jan 7 23:23:49 pfsense ntpd[86455]: 192.168.1.10 905a 8a sys_peer Jan 7 23:44:03 pfsense ntpd[86455]: 0.0.0.0 0613 03 spike_detect -0.145844 s Jan 7 23:53:04 pfsense ntpd[86455]: 0.0.0.0 0615 05 clock_sync Jan 8 06:15:15 pfsense ntpd[86455]: 0.0.0.0 0613 03 spike_detect -0.128529 s Jan 8 06:41:54 pfsense ntpd[86455]: 0.0.0.0 0615 05 clock_sync Jan 8 14:00:02 pfsense ntpd[86455]: 0.0.0.0 0613 03 spike_detect -0.130852 s Jan 8 14:08:35 pfsense ntpd[86455]: 0.0.0.0 0615 05 clock_sync
Updated by Kill Bill over 7 years ago
Try with TSC or TSC-low. Anyway, this ain't a pfSense bug, needs to go upstream.
Updated by Jim Thompson over 7 years ago
- Status changed from New to Not a Bug
OpenNTPD isn't better, it's worse.
Updated by John Silva over 7 years ago
Trying with TSC-low now. That seems to have done the trick! The time offset is staying relatively stable now. That was the one option that didn't really appear in any of the "Internet wisdom" on the topic. Thanks, @Kill Bill!