[26755] in Athena Bugs
Re: sunblade 150s crash with corrupt gconf
daemon@ATHENA.MIT.EDU (John Hawkinson)
Tue Feb 28 16:47:44 2006
Date: Tue, 28 Feb 2006 16:46:59 -0500
From: John Hawkinson <jhawk@mit.edu>
To: Robert Basch <rbasch@mit.edu>
Message-ID: <20060228214659.GP26428@multics.mit.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <EC1026C6-E1C3-4E70-B83D-98A25C4217F6@mit.edu>
X-Spam-Score: 1.217
X-Spam-Level: * (1.217)
X-Spam-Flag: NO
Cc: bugs@mit.edu
Errors-To: bugs-bounces@mit.edu
Robert Basch <rbasch@MIT.EDU> wrote on Tue, 28 Feb 2006
at 16:37:14 -0500 in <EC1026C6-E1C3-4E70-B83D-98A25C4217F6@mit.edu>:
> Hmm, I don't think anything in the boot sequence should be looking
> at /etc/gconf. (We do recreate gconf schemas in
> /etc/rc2.d/S90athena- ws, but that should only look at
> /etc/athena/gconf).
Well, gconftool-2 gets run from athena-ws. When I debugged this the
first time, gconftool hung, so of course the first thing I did was
move those files out of the way. (my metric was boot kadb, and
then ::ptree while it was hanging).
> What does the broken /etc/gconf look like?
I don't recall. You could go and look, though...
[I mean, I could too, and I understand you work in W92,
but I'm not directly in front of these machines, either...]
I kind of shudder at gconf and my immediate reaction is to hide from
it as much as possible and treat it like an opaque ball of
nastyness...
> Do you know if the machine hangs again if you move the broken
> /etc/gconf back into place?
I do not.
> With the advent of SMF in Solaris 10, this output is now logged to
> files in /var/svc/log; the output from /etc/rc2.d scripts will be in
> /var/svc/log/milestone-multi-user:default.log.
Good to know, thanks.
> In this situation, it might be useful to try doing "boot -m verbose",
> which prints messages as each service is started. "boot -m debug"
> will print exhaustive debugging info from SMF.
This is difference from boot -v, I guess, since I had tried that?
It would be *really* nice to get the output to go to the console as
well, so people (esp! those who are not all too knowledgeable about
solaris internals) could have an idea of what is ging wrong and the
ability to produce a useful bug report, rather than just saying, "this
machine is broken" (which is what seems to've happened in both the
cases I found this...).
Thanks.
--jhawk
