On Jun 18, 2016, at 6:37 AM, Embedding Linux <embed...@ba-cst.net> wrote: > We have a bunch of net6501 boxes around here. The hardware clock suffers > from quite high clock drift (4 to 10 secs a day, depending on the box). > > We are using NTP, so the soft clock is in sync (this is how we > discovered the hw clock drift). But when the boxes reboot, the soft > clock starts from the hw clock, and getting back to sync takes time (up > to several days).
You didn’t mention what OS you’re using (although I suppose your e-mail handle is a clue). However it is customary to write the system time to the hardware clock on shutdown, assuming you’re doing a graceful shutdown/reboot. Otherwise look at adding a cron job to do the same thing on a periodic basis. I would also look at doing an initial “ntpd -q -G” (formerly ntpdate) on boot, before starting ntpd normally. Which should make the value of the h/w clock irrelevant, except for early boot logging. And even then, ntpd will log how far the time is adjusted, so you could reconstruct accurate time for the early log values if this is critical. > Is such a big hw clock drift usual on 6501 ? Can’t answer for the 6501 specifically, but PC time of day clock crystals have traditionally had a huge variation in quality. And I don’t think any of them are temperature compensated. But it’s been a long time since I read up on this, so YMMV. Of course being a h/w geek, I added a Garmin GPS puck to my 5501 to keep time. -Jed _______________________________________________ Soekris-tech mailing list Soekris-tech@lists.soekris.com http://lists.soekris.com/mailman/listinfo/soekris-tech
