On Jan 5, 2005, at 1:36 AM, David N. Lombard wrote:
If there is a current default, but it doesn't work, and pool.ntp.org does work, we could make the latter the default. If neither works, I would leave the current value as the default -- it's certainly no worse than pool.ntp.org.
We should define "works" as "allows ntp two operate."
Hmm. Not sure I agree with that.
ntp will operate properly even if there is no valid server (i.e., it'll just keep the head node and client nodes in sync). And while that's useful, I don't know if it meets the law of least surprise. Yes, your cluster will be ok, and NFS will be nicely kept in sync, but what if your timestamps are 2 days in the future? (or, more realistically, several hours off from real time)
In this case, mail from your cluster (e.g., admin or root-level notices, or PBS/Torque notices) will be timestamped wrong. In a best case scenario, upon receipt of a mail from your cluster (assumedly at your workstation/mail server outside of the OSCAR cluster), you won't know it's real origination time unless you login to the cluster and manually compute the delta between real time and cluster time. In a worst case scenario, mail coming from the cluster will be treated as spam (and potentially discarded!) because the timestamps are so wildly wrong.
So I think we should define "works" as "properly synchronizes to a higher stratum server" -- *unless* the user specifically doesn't want it to synchronize to the outside world (i.e., if "none" is selected in the configurator).
Just my $0.02...
--
{+} Jeff Squyres
{+} [EMAIL PROTECTED]
{+} http://www.lam-mpi.org/------------------------------------------------------- The SF.Net email is sponsored by: Beat the post-holiday blues Get a FREE limited edition SourceForge.net t-shirt from ThinkGeek. It's fun and FREE -- well, almost....http://www.thinkgeek.com/sfshirt _______________________________________________ Oscar-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/oscar-devel
