Re: Speaking of ntp, yes CD-R does mess with the clock!
On Tue, 24 Dec 2002, Rob Weir wrote:
> I'd say this is the sort of thing people mean when people say 'IDE
> sucks!'; the CPU has to babysit the burner through the entire process...
It's hard for me to believe that can be the only problem. The clock falls
behind almost a minute during a 4 minute burn session. Yet, during the
burn I can still run other applications and the machine seems to respond
normally. The machine is a PIII450 with a 48x CD-R drive. The machien is
not a screamer, but plenty fast.
Although I don't use them much, I have two other machines with ide-scsi
CD-Rs (16x burn speed) that do not get behind at all.
I suspect that there must be another issue, otherwise a lot more people
would be asking about this. I just need to sit down with the machine and
burn a bunch of coasters and watch what happens.
There was a power failure where that machine is yesterday -- about an hour
or so without power. I just checked and right now the system clock and
hwclock is *ahead* of real time by about 45 seconds. Clearly ntpd is not
working.
ntp's problem is it give up too easily. This is a pppoe connection.
/etc/rc2.d/S14ppp brings up the connection, and ntp starts with the S23
script. syslog shows output from pppd before ntpd starts, but also right
*after* ntpd fails to connect. So ntpd is starting too soon. But it's
bad that ntp doesn't every try again after that first failure.
I just don't think it unreasonable that there could be periods of time
when ntp can't connect to the remote hosts -- and that should not stop
ntp.
--
Bill Moseley moseley@hank.org
Reply to: