Re: Debugging time

Victor Snezhko wrote:
> Ivan Voras <ivoras_at_fer.hr> writes:
> 
>>> For about half a year I have another weirdness with time on my
>>> -current box without vmware, maybe it's somehow has common origins
>>> with your behaviour. When I boot FreeBSD, ntpdate (which is set up to
>>> run via rc.conf) often reports huge time offsets:
>>>   10 Jul 12:04:34 ntpdate[51761]: step time server ntp.server.name offset 234456.074166 sec
>>>
>>> I have so far only noticed that the value of the the offset is roughly
>>> equal to the amount of time when FreeBSD isn't running (this is a
>>> dual-boot box with windows). The box is pretty old, 600-MHz Celeron
>> Is the above number correct (i.e. ~~ 65 hours)? 
> 
> Yes, sometimes my FreeBSD is not running for several days - in this
> case ntpdate prints such numbers. It looks like kernel stores date and
> time values somewhere and then re-reads them. I still don't rule out
> some stupidity on my part - the behaviour I'm observing is very
> illogic. Maybe I should try installing current from a fresh snapshot
> when it's released.
> 
>> I noticed there can be a consistent "shift" of duration equal to
>> your timezone when VMWare sets the RTC to the host's local time and
>> FreeBSD thinks it's UTC. But if the difference is not constant, it's
>> not it.
> 
> When I just reboot FreeBSD, offset doesn't exceed one or two seconds
> (notably, bios takes longer to boot, about 5-10 seconds). When BSD is
> inactive for several hours, number is much bigger.
> 
> I neglected to debug this deeply before - I assumed I have overlooked
> something simple, but now that you have reported another issue in the
> similar area, I'll try to collect full information.
> 


When you say 'inactive', do you mean that you have suspended the machine 
in vmware, or it is 'powered off'?

Eric