[261] in athena10
Re: Athena 10 on ashdown04 cluster machine: installation report and
daemon@ATHENA.MIT.EDU (Kenneth Charles Arnold)
Sun Jun 22 00:24:05 2008
Message-ID: <485DD3B6.5040903@mit.edu>
Date: Sun, 22 Jun 2008 00:23:18 -0400
From: Kenneth Charles Arnold <kcarnold@MIT.EDU>
MIME-Version: 1.0
To: Timothy G Abbott <tabbott@mit.edu>
CC: ashdown-webmaster@mit.edu, athena10@mit.edu
In-Reply-To: <alpine.DEB.1.10.0806211737110.14493@vinegar-pot.mit.edu>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Timothy G Abbott wrote:
> Perhaps `hostname` is not returning ashdown04? Unless you mean that
> CUPS doesn't know the default printer, which I would not find surprising.
PEBKAC. The Ubuntu installer suggested ASHDOWN-something, and I just
removed the something and left ASHDOWN04. Changed to lower case and it
works fine now.
> These are because the 'mit' group has gid 101, which conflicts with
> local groups (I guess libuuid got 101 on your system). On Linerva we
> manually renumber the local group with gid 101 to some other number.
# find / -mount -group libuuid
/var/lib/libuuid
/usr/sbin/uuidd
So it would not be too invasive to do said manual renumbering. But
definitely not a good long-term solution :)
> The right long-term solution to this problem is probably for the 'mit'
> group to find a new number in the ranges that aren't used by local
> systems, but I don't see how to get there from here.
I'm not sure how hesiod works, but maybe it would be possible for 101 to
map to 'mit' and 'mit' map to some higher number? Then a gradual
conditional chgrp on the AFS servers wouldn't be intrusive.
Other notes from this machine:
1. Ripped out cups. replaced by (non-Debathena) lprng (killed
bluez-cups, cups-pdf, foomatic-db-hpijs, hpijs, hplip, hal-cups-utils,
cupsys-driver-gutenprint, cupsys, cupsys-client). Maybe that was a
little excessive. But printing works now. (cups is still preferred, but
now it can wait.)
2. After adding the evolution wrapper, it still says on startup:
---
Upgrade from previous version failed: Failed upgrading Mail settings or
folders.
Unable to read settings from previous Evolution install,
`evolution/config.xmldb' does not exist or is corrupt.
If you choose to continue, you may not have access to some of your old data.
---
It then proceeds to the initial setup wizard. Less than helpful. Is it
too much of a version jump to do at once?
3. To tabbott's concern about openssh-server, that's why I disabled root
login in the config file. The openssh-server package itself is hard to
remove, since debathena-workstation Depends debathena-login Depends
debathena-ssh-server-config Depends openssh-server. :(
4. The ssh client problem mentioned a few weeks ago is still around.
Fixed by setting:
GSSAPIDelegateCredentials yes
in /etc/ssh/ssh_config.
5. I had to symlink my .thunderbird to .mozilla-thunderbird to make
thunderbird use my profile. Seems like somebody moved something. Any
estimate on how many people use Thunderbird?
6. The logout window was faster this time. Don't know what it was before.
7. Installed sun-java6-jre (agreed license agreement...), export
MATLAB_JAVA=/usr/lib/jvm/java-6-sun-1.6.0.06/jre, fixes MATLAB. I put
that in /etc/environment for now; probably a good idea to use local Java
in general. Ubuntu bug page reference:
https://bugs.launchpad.net/ubuntu/intrepid/+source/libxcb/+bug/185311 --
in short, seems like the new libxcb stuff is less tolerant of sloppy X
client code. We should check other proprietary apps for similar issues
(though I'm not familiar enough with any of them to do that myself).
8. Firefox was occasionally freezing up, especially clicking things on
Youtube. Disabling disk cache seems to fix that.
9. The dotfiles don't source bash_completion. I miss it
10. I think Firefox has some startup lock issues. If you click the
launch a few times, it complains about a lockfile but adds another tab
to an existing session. And after I quit it, the binary is still sitting
around in a FUTEX_WAIT call, locking up X resources (see xrestop and get
enlightened).
11. Overall speed is not great. Is there any systematic way to audit
network access? (It's a decent machine, just on a 10 Mb network. But I'm
seeing lags on things that I don't think should need a network round-trip.)
-Ken
