Asktime and ntp problem on 5.0.7
From: Steve M. Fabac, Jr. (smfabac_at_att.net)
Date: 01/21/05
- Next message: moncho: "talkd"
- Previous message: Bela Lubkin: "Re: Problem with Streams someone please help."
- Next in thread: Bob Bailin: "Re: Asktime and ntp problem on 5.0.7"
- Reply: Bob Bailin: "Re: Asktime and ntp problem on 5.0.7"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Fri, 21 Jan 2005 19:05:39 GMT
I've noticed that ntp and asktime have problems on 5.0.7.
I set up /etc/ntp.conf as follows:
> #
> server time-a.nist.gov
> #server tick.usno.navy.mil
> #server time-b.nist.gov
> #server time.nist.gov
> #server time-a.timefreq.bldrdoc.gov
>
> # ^ The IP address of the stratum 0 server machine that you would
> # be asking for the time.
>
> # Declare local clock as stratum 10 server in case we loose contact with
> # Strata 1 servers and so that our machine will serve time to the network.
>
> server 127.127.1.0
> fudge 127.127.1.0 stratum 10
>
> driftfile /etc/driftfile
>
> # ^ A scratch file used by xntpd for time adjustment
Then I modify /etc/tcp as follows to run ntpdate upon boot up.
This is the same modifications I have been using on 5.0.2 - 5.0.6
without problems.
> #
> # ntpd - Don't run this at the same time as timed.
> if [ -f /etc/ntp.conf ]; then
> # Nb. tickadj is NOT necessary or recommended.
> primary=`grep server /etc/ntp.conf | head -1 | awk ' { print $2 } ' `
> ntpdate $primary
> /etc/ntpd -g
> echo "ntpd \c"
> fi
>
> #
This requires that the default route be set up prior to ntpdate running
above as it won't work if it can't get off the machine. Prior to 5.0.7,
I add a route add default directive in /etc/tcp after the
/etc/route add 224.0.0.0 192.168.10.1 0 > /dev/null 2>&1
command. 5.0.7 sets the default route prior to executing the ntp section.
> ntpdc> peers
> remote local st poll reach delay offset disp
> =======================================================================
> =LOCAL(0) 0.0.0.0 10 64 377 0.00000 0.000000 0.00095
> *time-a.nist.gov 192.168.10.1 1 1024 377 0.08694 0.008310 0.02217
> ntpdc>
The problem: On the two new 5.0.7 system I set up, the ntpdate command
will indicate that the internal clock is off by 20000+ seconds and make
the adjustment ( set -x in /etc/tcp and watch the boot up hardware
enumeration screen). Ntp takes care of the problem and executing date
returns the correct time and date. If you enter the correct time at the
boot up prompt, the hardware clock is set to the wrong time and after
several quick boot up cycles can be days off. Setting the BIOS hardware clock
to the correct date and not touching the keyboard when the system boots
so that you don't respond to asktime, usually allows the system to boot
without the 20000+ second offset being announced.
This is a rambling description of the problem, as I have delivered the
servers and am not trying to solve the problem, just get this out to the
community to see if anyone can test this on their systems and verify or
report that everything works as expected. (In which case, I have to assume
that the SuperMicro X5DL8-GG system board is the problem.)
--
Steve Fabac
S.M. Fabac & Associates
816/765-1670
- Next message: moncho: "talkd"
- Previous message: Bela Lubkin: "Re: Problem with Streams someone please help."
- Next in thread: Bob Bailin: "Re: Asktime and ntp problem on 5.0.7"
- Reply: Bob Bailin: "Re: Asktime and ntp problem on 5.0.7"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
