On 05/02/17 16:15, Richard wrote:> >> Date: Sunday, February 05, 2017 10:26:05 -0500 >> From: Robert Moskowitz <rgm at htt-consult.com> >> >> I have read: >> http://thegeekdiary.com/centos-rhel-7-chrony-vs-ntp-differences-bet >> ween-ntpd-and-chronyd/ >> >> My server is up all the time and will serve time to internal >> systems (via DHCP options). >> >> Caveat is that my server is an armv7 (Cubieboard2) which does not >> have an RTC (no battery). So whenever the system boots, the time >> is ZERO (Dec 31, 1969 or some such). >> >> Chrony fixes this really fast; shortly after boot the time is good. >> Chrony CAN be configed as an internal time server. But chrony does >> not seem to step the clock for any adjustments needed. It is more >> important that this systems time be right all the time than to >> avoid clock steps. >> >> This brings me back to NTP, which normally takes hours to bring the >> time from ZERO to current, but keeps the time correct. >> >> So: >> >> Can Chrony check the time, say once a day? >> >> Or can NTP make a BIG time jump all at once (on system restart)? > > Where I have somewhat similar issues, I have historically used a > crontab "@reboot" entry to call ntpdate which gets the clock set > correctly. From there ntp keeps it in sync. > > This can now be accomplished with ntpd, and ntpdate is threatened > with depreciation/retirement. See the top of the ntpdate man page for > more details. >The NTP configuration option you may be after is "tinker panic 0" which allows NTP to make big jumps as often as required. See ntp_misc(5). There is a related discussion with making VMs take big jumps at https://www.centos.org/forums/viewtopic.php?f=16&t=61186&p=258254#p258254 -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 836 bytes Desc: OpenPGP digital signature URL: <http://lists.centos.org/pipermail/centos/attachments/20170205/5cad9399/attachment-0001.sig>
On 02/05/2017 11:58 AM, J Martin Rushton wrote:> On 05/02/17 16:15, Richard wrote: >>> Date: Sunday, February 05, 2017 10:26:05 -0500 >>> From: Robert Moskowitz <rgm at htt-consult.com> >>> >>> I have read: >>> http://thegeekdiary.com/centos-rhel-7-chrony-vs-ntp-differences-bet >>> ween-ntpd-and-chronyd/ >>> >>> My server is up all the time and will serve time to internal >>> systems (via DHCP options). >>> >>> Caveat is that my server is an armv7 (Cubieboard2) which does not >>> have an RTC (no battery). So whenever the system boots, the time >>> is ZERO (Dec 31, 1969 or some such). >>> >>> Chrony fixes this really fast; shortly after boot the time is good. >>> Chrony CAN be configed as an internal time server. But chrony does >>> not seem to step the clock for any adjustments needed. It is more >>> important that this systems time be right all the time than to >>> avoid clock steps. >>> >>> This brings me back to NTP, which normally takes hours to bring the >>> time from ZERO to current, but keeps the time correct. >>> >>> So: >>> >>> Can Chrony check the time, say once a day? >>> >>> Or can NTP make a BIG time jump all at once (on system restart)? >> Where I have somewhat similar issues, I have historically used a >> crontab "@reboot" entry to call ntpdate which gets the clock set >> correctly. From there ntp keeps it in sync. >> >> This can now be accomplished with ntpd, and ntpdate is threatened >> with depreciation/retirement. See the top of the ntpdate man page for >> more details. >> > The NTP configuration option you may be after is "tinker panic 0" which > allows NTP to make big jumps as often as required. See ntp_misc(5). > There is a related discussion with making VMs take big jumps at > https://www.centos.org/forums/viewtopic.php?f=16&t=61186&p=258254#p258254Thanks. I will look at this. All I was seeing was to use burst and iburst, but they would not make the really big jump needed after boot.
On Sun, 2017-02-05 at 12:30 -0500, Robert Moskowitz wrote:> > On 02/05/2017 11:58 AM, J Martin Rushton wrote: > > On 05/02/17 16:15, Richard wrote: > > > > Date: Sunday, February 05, 2017 10:26:05 -0500 > > > > From: Robert Moskowitz <rgm at htt-consult.com> > > > > > > > > I have read: > > > > http://thegeekdiary.com/centos-rhel-7-chrony-vs-ntp-differences > > > > -bet > > > > ween-ntpd-and-chronyd/ > > > > > > > > My server is up all the time and will serve time to internal > > > > systems (via DHCP options). > > > > > > > > Caveat is that my server is an armv7 (Cubieboard2) which does > > > > not > > > > have an RTC (no battery).??So whenever the system boots, the > > > > time > > > > is ZERO (Dec 31, 1969 or some such). > > > > > > > > Chrony fixes this really fast; shortly after boot the time is > > > > good. > > > > Chrony CAN be configed as an internal time server.??But chrony > > > > does > > > > not seem to step the clock for any adjustments needed.??It is > > > > more > > > > important that this systems time be right all the time than to > > > > avoid clock steps. > > > > > > > > This brings me back to NTP, which normally takes hours to bring > > > > the > > > > time from ZERO to current, but keeps the time correct. > > > > > > > > So: > > > > > > > > Can Chrony check the time, say once a day? > > > > > > > > Or can NTP make a BIG time jump all at once (on system > > > > restart)? > > > > > > Where I have somewhat similar issues, I have historically used a > > > crontab "@reboot" entry to call ntpdate which gets the clock set > > > correctly. From there ntp keeps it in sync. > > > > > > This can now be accomplished with ntpd, and ntpdate is threatened > > > with depreciation/retirement. See the top of the ntpdate man page > > > for > > > more details. > > > > > > > The NTP configuration option you may be after is "tinker panic 0" > > which > > allows NTP to make big jumps as often as required.??See > > ntp_misc(5). > > There is a related discussion with making VMs take big jumps at > > https://www.centos.org/forums/viewtopic.php?f=16&t=61186&p=258254#p > > 258254 > > Thanks.??I will look at this.??All I was seeing was to use burst and? > iburst, but they would not make the really big jump needed after > boot. > >With NTP you could use the ntpdate.service as well as/ before ntpd.service. the former is supposed to set the clock once BEFORE ntpd is started. See: https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_L inux/6/html/Deployment_Guide/s1-Configure_ntpdate_Servers.html Ntpdate seems to work on RHEL/Centos 7 as well... /Louis