[mdlug] 2 very different gigabyte motherboards not keeping good time

[Dan Pritts]

Thanks for taking the time to reply.  

All this said, I started NTP again on monday, and the system has kept proper time.  

I suppose it’s possible that somehow I had an out-of-sync clock that wasn’t set 
properly at boot time, and ntp hadn’t been able to catch up yet.  But, ntp starts 
up with the -g option so that seems unlikely too.  Maybe the network didn’t 
come up properly at first, or something.  

Piss me off.

On Apr 15, 2014, at 6:07 AM, Adam Tauno Williams <awilliam at whitemice.org> wrote:

> On Mon, 2014-04-14 at 22:32 -0400, Dan Pritts wrote:
>> For several years I’ve run an old 1.9GHz 64-bit athlon, with a gigabyte motherboard.  
>> It never kept good time, whether running under freeBSD (8.x then 9.x) or linux (centos6).  
>> NTP couldn’t keep it synced, 
> 
> In 25 years as a sys-admin I've never met a box that NTP couldn't keep
> in sync - at least not one where the settings weren't completely
> wrong.  

I’ve only been at it for 20 years, but I haven’t had this problem elsewhere either.  Thus
my confusion.  Here’s my ntp.conf. 

I’ve diddled a little - i added tinker panic 0 in an attempt to keep NTP from giving up.    
But I don’t think it’s givng up, I think it just can’t keep up.   I think there is an option that 
lets it make larger corrections to the clock, “steps” instead of “slews”.   I think I tried 
tinkering with this somewhere in the past but I don’t remember the details.   It didn’t work
well and I gave up and put in an ntpdate cron job.  


~@kompressor% grep -v '^#' /etc/ntp.conf | sort -u
driftfile /var/lib/ntp/drift
includefile /etc/ntp/crypto/pw
keys /etc/ntp/keys
restrict 127.0.0.1
restrict 207.75.164.74 mask 255.255.255.255 nomodify notrap
restrict -6 ::1
restrict -6 default kod nomodify notrap nopeer noquery
restrict default kod nomodify notrap nopeer noquery
server 141.211.125.15 iburst
server 207.75.164.74 iburst minpoll 8
server 2.us.pool.ntp.org minpoll 8
server 3.us.pool.ntp.org minpoll 8
tinker panic 0





> The clock drifts significantly WHILE-THE-BOX-IS-RUNNING *OR* is totally
> kitty-whumpus AFTER-RESTART/BOOT?

While the box is running.

>> I recently upgraded the system to a “pentium” haswell chip. 
>> Guess what, it doesn’t keep time, and NTP can’t keep up.
> 
> And what components did you retain between the two systems?  Video card?
> Some other device?  

disk drives and a PCI-express SATA controller.  The old system didn’t work right before 
the SATA controller was added.  

Also a PS/2 keyboard, and a VGA monitor.  

Same LAN and upstream comcast connection.  Comcast modem was replaced
about a year ago.  

Running ZFS, first native under FreeBSD, now with zfsonlinux.org packages.

> You have a daisy chained mess of USB devices hanging off the box which
> resembles a rail switching yard?

None

> You diddle with the kernel or set some unusual boot-string options
> regarding APIC or ACPI?

Nope

>> I do not have the system on a UPS or anything else that might muck with 
>> the power frequency coming in.  Just plugged in to the wall, 
>> DTE power in the middle of ann arbor.
> 
> There is no circuit which references the 'socket clock' in a modern box.

I would expect htere is some sort of crystal oscillator that handles the clock,
but then, I would expect it would keep good time too, so what do i know.

danno