[27545] in Athena Bugs
Re: debathena machines don't notice when they're off the network
daemon@ATHENA.MIT.EDU (Jonathan Reed)
Sun Apr 3 20:07:12 2011
Mime-Version: 1.0 (Apple Message framework v1084)
Content-Type: text/plain; charset=us-ascii
From: Jonathan Reed <jdreed@mit.edu>
In-Reply-To: <201104032338.p33NcLtS029786@multics.mit.edu>
Date: Sun, 3 Apr 2011 20:07:08 -0400
Message-Id: <D060834B-D2C0-46C4-A450-684977AE8F28@mit.edu>
To: John Hawkinson <jhawk@mit.edu>
Content-Transfer-Encoding: 8bit
Cc: bugs@mit.edu
Errors-To: bugs-bounces@mit.edu
Debathena machines have several mechanisms by which they notice when they're off the network. They will report "Cannot contact the Athena login servers" via PAM (which shows up at the bottom of the login dialog) if they can't reach the KDC when attempting to get the TGT. (I know this still works, as I encountered it in W20 2 weeks ago when a machine's network cable was unplugged). We also have an open ticket for machines warning the user when their AFS is homedir is unavailable, which is currently in testing in the alpha cluster.
Now, neither of these will be effective when the network is just slow, or fails in other spectacular manners, but I don't really have a good answer there. Your description of the machine taking a few seconds to produce a username prompt (presumably this delay is between clicking the "Log In" button and the username field appearing) is baffling, because there should be nothing network-related going at that time.
-Jon
On Apr 3, 2011, at 7:38 PM, John Hawkinson wrote:
> Currently the Bldg. 12 cluster seems to be off the network, reported as
> [help.mit.edu #1529813]. Trying to login on those machines is unfortunate --
> they don't give any sign that they are off-net. Instead, they poke along:
>
> Takes a few seconds to produce a username prompt.
> Takes ~30 seconds to produce a password prompt.
> Sit there for minutes (I waited 3 or 4) trying to login.
>
>
> It would be nice if there was a clear indication that something was wrong --
> a "please don't use me," etc. Surely the machine should know when the
> hesiod lookup fails, presuambly before it gets to the password prompt?
>
> Thanks.
>
> --jhawk
