On Fri, 23 Sep 2005, Matthew Burgess wrote:
Ken Moffat wrote:
I think it depends where you go for your timeservers - on my server
(normally up, modulo hardware changes and kernel crashes) I've commented
out the 'ntpd -gqx' with a note that I was getting an 'already running
message' (gcc-3.4.3, ntp-4.2.0). Works fine like this.
That 'already running' message is a symptom of the ntpdate-esque ntpd not
having finished running by the time the sync-mode ntpd gets kicked off. Or
at least that's what I concluded when seeing the same message here during my
own tests.
Agreed, but I don't know why, and my current setup works well enough
for me not to care to investigate. I'm assuming your box hadn't been
running on local time, and the time in the bios was reasonably accurate.
As for the timeservers, I've tested with just my router and with a
combination of the ones listed in the BLFS book (generally just using the
European one though to save time).
My servers predate that addition to the book, although they're not what
I can class as really local: bear.zoo.bt.co.uk, time-server.ndo.com,
ntp2b.mcc.ac.uk. There was a better (nearer) server, but it went away.
Unfortunately, ntl isn't clued up enough to run a server for its
customers. Using international servers is undoubtedly a fascinating
proof of how wonderful the internet is, but nearer is usually better.
Can't remember where I found these (somewhere in or linked from the
online ntp docs), but I can remember spending some time pinging public
servers before trying to use a few of them.
On my desktops/laptop (which get ntp from my server) the ntpd -gqx works
fine after bios resets (or ibook system resets) and results in only a short
delay during any boot. No detectable difference between
gcc-3.3.5/gcc-3.4.3/gcc-4.0.1 on the desktops (all with ntp-4.2.0, gcc4
with the patch).
So, just to confirm: Your ntp bootscript is unchanged on these machines (and
is from a release newer than October-2004), you have them syncing to remote
timeservers and those machines complete the ntp bootscript in, say, < 2
seconds? Do they have anything particularly quirky in the /etc/ntp.conf
file?
No, I have them syncing to my server, the only thing in ntp.conf on
the desktops is 'server deepthought'. That's the only box syncing to
remote timeservers [ ignoring the firewall/router, which is still using
chrony ]. They aren't _that_ quick to complete the bootscript, it feels
more like five seconds. The ntp bootscripts are all from this year.
Of course, it's possible this is a problem with 4.2.0a, or perhaps you
upgraded something else (e.g. the kernel) ?
Hmmm, good point, I'm actually on 2.6.13.2 here. I spotted a LKML post
(http://www.ussg.iu.edu/hypermail/linux/kernel/0509.2/0514.html) suggesting
it might be broken, but noone can apparently confirm or deny that.
2.6.13 works as well, or badly, as 2.6.12.x on a desktop here (on
amd64, as previously reported by someone, clock is several seconds
adrift after a reboot).
I'll try
reverting to ntp-4.2.0 then to linux-2.6.12.6 to see if they fix the problem.
Regards,
Matt.
Sorry for the delay responding, I'm still struggling with a cold - you
can get quicker answers if I know it's important, but they may not be
grounded in the real world ;)
Ken
--
das eine Mal als Tragödie, das andere Mal als Farce
--
http://linuxfromscratch.org/mailman/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
