[OT] NTP servers - was Re: Kernel clock issue "Clocksource tsc unstable"

[Brian McKee]

On Sat, Feb 28, 2009 at 9:06 AM, Derek Broughton <derek at pointerstop.ca> wrote:
> Brian McKee wrote:
>
>> On Fri, Feb 27, 2009 at 11:22 PM, Rashkae <ubuntu at tigershaunt.com> wrote:
>>> NoOp wrote:
>>>
>>>>
>>>> I also purposely remove the ntp.ubuntu.com/canonical servers; I' don't
>>>> mind providing some data, but sync'ing to them just seems a little big
>>>> brother/Microsoft'ish to me.
>>
>> as you wish - but I don't recall ntp sending any data the other way
>> worth mentioning...
>
> The fact that you continually connect is data.  I can see his point - though
> _I_ remove ntp.ubuntu.com because I've found it notoriously hard to
> reach...

It's hardly continual - and I connect to my local gateway, not ubuntu,
but whatever.

>
>>> Unless you're doing something that's extremely time sensitive (such as
>>> synchronizing to those time based RSA security tokens) I don't really
>>> see the point to synchronizing with ntp at all.   Any functioning clock
>>> should be able to keep track of wall time within a few seconds accuracy
>>> per month.
>>>
>> You must have better quality systems than some of mine then - I've got
>> units with good batteries that wander several minutes out of whack in
>> a month without ntp all the time.

> A few seconds/several minutes.  Either way, it really isn't a big enough
> deal to worry about, usually, but then I prefer to set my computers to
> display analog clocks, because I don't like being too precise...

But I like my backups to work correctly, and my logs to be accurate,
so I can see what's going on.  The VPN I use for work doesn't like it
if your time is to far out of what either.

>
> otoh, it seems to me that "ntp all the time" and "units with good batteries"
> are conflicting criteria.  If your system is _capable_ of getting the time
> via ntp, then the clock is not running on the battery.  The clock you're
> seeing is the software clock, so if it's wandering while you're powered up
> and not using ntp it's got nothing to do with the CMOS battery (and the hw
> clock gets set from the sw clock on shutdown).  If the clock only wanders
> while you're powered off, then setting the sw clock at restart time via
> ntpdate would be sufficient.

Who cares when it wanders - on or off - I'd rather send one lousy UDP
packet every once in a blue moon to my own router and keep everything
actually at the right time.  I just don't see why you WOULDN'T do it!

Brian