[20202] in Athena Bugs
Re: LIDS Athena Cluster
daemon@ATHENA.MIT.EDU (Cana Lynn McCoy)
Thu Feb 28 18:19:17 2002
Date: Thu, 28 Feb 2002 18:19:16 -0500 (EST)
From: Cana Lynn McCoy <cana@MIT.EDU>
To: "Petr M. Swedock" <petr@mit.edu>
cc: <bugs@mit.edu>
In-Reply-To: <200202261358.IAA22611@nerd-xing.mit.edu>
Message-ID: <Pine.GSO.4.30L.0202281815100.2995-100000@pullman-standard.mit.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
I'm forwarding this from the sysadmin of the LIDS department, which has a
small cluster of privately maintained Athena machines.
On Tue, 26 Feb 2002, Petr M. Swedock wrote:
> Also, in the cluster there is a misconfiguration in the startup
> scripts: specifically, there are two calls to two different binaries
> that serve ntp. In the /etc/rc2.d/ directory there is a call to
> xntpd in /etc/rc2.d/S74xntpd and a call to ntpd in
> /etc/rc2.d/S90athena. I've seen this before on likewise misconfigured
> Sun boxen. This is bad, as often the first instance of ntpd
> will open the socket, while the second instance might fail to
> open the socket but not fail altogether... leaving both binaries to
> make changes to time_t (or whatever the function is, I forget)
> while only one is actually syncing to a network clock. It's not
> gonna hurt the kernel or anything, but my concern is for users
> running, say, Matlab across MITnet... They need to sync time up
> correctly. Or Kerboros, which (if I understand correctly) won't
> authenticate if machine time is significantly different from the
> kerberos realm.
>
> I don't know if this misconfiguration is only in the LIDS Athena
> cluster or more widespread. But I'd like to change it, or have
> it changed, in the LIDS cluster.
>
> Many thanks for your time and attention to this.
>
> Peace,
>
> Petr Swedock, LIDS
> ------- End of resent message -------
