-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of
Michael Edwards
Sent: Wednesday, February 16, 2005 16:53
To: OSCAR-DEVEL
Subject: Re: [Oscar-devel] [ oscar-Bugs-845179 ] ntpconfig
assumes internet connectivity
Incidently, when using the "localhost as the only ntp server"
solution, it reads as [Failed] on bootup (at least on my FC2 setup).
I have not investigated beyond checking ntp.conf, which looks right.
The clocks are all within a minute of each other anyway, but
I presume they should be within a second (or so).
On the plus side, it does fail out a ton faster.
On Wed, 16 Feb 2005 15:42:49 -0800, SourceForge.net
<[EMAIL PROTECTED]> wrote:
Bugs item #845179, was opened at 2003-11-19 07:12 Message generated
for change (Comment added) made by jcsiadal You can respond by
visiting:
https://sourceforge.net/tracker/?func=detail&atid=109368&aid=845179&gr
oup_id=9368
Category: Installation
Group: 4.1
Status: Closed
Resolution: Duplicate
Priority: 5
Submitted By: Jeff Squyres (jsquyres)
Assigned to: Bernard Li (bernardli)
Summary: ntpconfig assumes internet connectivity
Initial Comment:
The install scripts for ntpconfig assume that the OSCAR cluster is
connected to the greater internet. For example, it assumes
that the
following public NTP servers can be reached:
ntp-1.cso.uiuc.edu
ntp0.cornell.edu
ntp.cmr.gov
If you are installing OSCAR on a disconnected network (say,
for a demo
at SC2003) where such machines are not available, it causes lengthy
timeouts during multiple phases of the OSCAR installation
(where ntp
is stopped/started and the presence of the
/etc/ntp/step-tickers file
makes ntp try to sync with an available time server).
The scripts should check for connectivity to these servers (e.g., a
DNS resolve and/or a ping) before assuming anything. Failing
connectivity, they should not be added to step-tickers file.
In a worst case scenario, only 127.0.0.1 should be in
step-tickers so
that the clients can time sync to the head node so that NFS
activity
will be kosher.
----------------------------------------------------------------------
Comment By: Jeremy Siadal (jcsiadal)
Date: 2005-02-16 15:42
Message:
Logged In: YES
user_id=847865
The local (system) clock for NTP is specified by the pseudo- IP
address 127.127.1.0.
Don't confuse it with the localhost address, unless ntpconfig takes
care of that conversion for us.
----------------------------------------------------------------------
Comment By: Bernard Li (bernardli)
Date: 2005-02-15 16:33
Message:
Logged In: YES
user_id=879102
Duplicate, see bug 1084640.
----------------------------------------------------------------------
Comment By: Jeff Squyres (jsquyres)
Date: 2003-11-22 05:43
Message:
Logged In: YES
user_id=11722
Everything you say makes perfect sense.
I guess my point is that it would be nice to be automatic.
Perhaps if ping fails to turn up a valid NTP server, then
perhaps we
can try an ntp sync with various servers and see if that works.
I understand that this would move the delay to the install process
(vs. the boot process), but this would actually be preferable,
particularly if the user can somehow be told what is happening
(probably initially via a log message in the text/output
window, but
eventually via a nice popup window saying "I'm looking for a
valid NTP server -- this may take a few seconds...").
----------------------------------------------------------------------
Comment By: Terry Fleury (tfleury)
Date: 2003-11-19 12:22
Message:
Logged In: YES
user_id=116599
Since you can't deselect ntpconfig for installation, this is rather
annoying. If I can add ntp to the base rpm list (so that you can
deselect ntp), the user could avoid the waits by deselecting
ntpconfig. ntp would still be installed, but ntp is not started by
default.
The other easy solution is to go into the Configurator for
ntpconfig,
uncheck the three default servers, and then enter
127.0.0.1 into the 'extra ntp servers' textbox. That was
one reason
ntpconfig was created in the first place.
Relying on some method to check for the presence of the
servers (such
as ping) is not guaranteed to give you correct results.
For example,
my home cable modem company has disabled pinging from home,
but I can
use the school's ntp server without any problem. I believe
there is
an ntp utility that allows you to check for the presence of a given
ntp server, but that requires that ntp be installed already.
Besides, checking for an ntpserver early on only moves the
delay you
see to an earlier time. The best/easiest solution I see now is to
either (a) not install ntpconfig (which will be able to
occur once ntp
is added to the base RPM list) or
(b) configure ntpconfig to use 127.0.0.1 as the only ntp server.
----------------------------------------------------------------------
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=109368&aid=845179&gr
oup_id=9368
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide Read honest
& candid
reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start
reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
_______________________________________________
Oscar-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/oscar-devel
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide Read honest &
candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
_______________________________________________
Oscar-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/oscar-devel
