[6463] in Release_7.7_team

home help back first fref pref prev next nref lref last post

Minutes from 10/9 meeting

daemon@ATHENA.MIT.EDU (Jonathan Reed)
Fri Oct 9 16:07:03 2009

Message-Id: <B9090C60-86FC-4E4F-9900-FC4D9F52F6FD@mit.edu>
From: Jonathan Reed <jdreed@MIT.EDU>
To: release-team@mit.edu
In-Reply-To: <BC524846-8079-4D5C-AC4C-A25925B482EB@mit.edu>
Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes
Content-Transfer-Encoding: 7bit
Mime-Version: 1.0 (Apple Message framework v919.2)
Date: Fri, 9 Oct 2009 16:06:55 -0400
X-Spam-Flag: NO
X-Spam-Score: 0.00

> I. Quickstation update

sudoers should get a message, they currently don't.  Removable devices  
need to be tested
We can deploy this next week (Tues). Known issue with updates under  
people's login sessions, FF may become unhappy.
Evan will double-check code to see if kernel updates cause a reboot.

ACTION ITEM:  jdreed will test removable devices and make sure they  
still work
ACTION ITEM:  broder/geofft will fix sudoers issue and double-check  
code to see what happens for kernel updates (though the second part is  
likely moot due to the following agenda item)

Target deployment date: Afternoon of Tue 10/13

> II. athena-auto-update consistency and failure modes

If athena-auto-update is interrupted, it gets sad and leaves packages  
unconfigured, which breaks things.  Hotline encountered this while  
force-logging out users in clusters.

Solution: athena-auto-update shouldn't run while users are logged  
in.   We will implement desync interval changes based on time of day.   
desync 2 hours from 2am - 8am, otherwise desync 6 hours.  When a  
machine is updating, we will touch /etc/nologin and we will switch a  
VT and display a message to users.   The desync intervals should  
prevent us from getting screwed by, e.g. texlive-full updates.

> III.  differences in public machines' package lists

There are differences.  Some of these are jdreed's fault.  We possibly  
don't care, but we want to implement some sort of public-workstation  
verification script, possibly involving debsums.  Discussion of  
technical details punted to zephyr.

> IV. Firefox speed proposal

Suggestions are to move the entire profile to tmpfs and sync back,  
move only the sqlite databases to tmpfs and sync, or do something with  
LD_PRELOAD.   This goes on the back burner, we're more concerned with  
FF startup times at the moment.

> V. Metrics update

Code is in proposed.   We should review the document in Hermes and  
publish it.   We need to provide ops with an aggregate script.

ACTION ITEM:  Evan will add client-side blacklist.
ACTION ITEM: jdreed will write a perl script for ops.
ACTION ITEM:  release-team will review https://kb.mit.edu/confluence/pages/viewpage.action?pageId=5376333

> VI. Migration to ops' CUPS servers

ACTION ITEM: jweiss will verify that we can start using  
printers.mit.edu for the cluster machines.
ACTION ITEM: broder will make cups config changes and move them into - 

We still need to come up with a solution for -workstation machines.   
Perhaps that's "Add the print queues you want by hand, like you do on  
Mac and Windows".

> VII. We really want to start using our own Ubuntu mirror

mirrors.mit.edu is not yet production according to the maintainers.   
We should set up our own mirror on athena10.mit.edu.   We may test  
with athena10 serving i386 jaunty only and see if that fixes our  
install problem.

ACTION ITEM: broder will provide jweiss with disk usage data for  
portions of an ubuntu repository
ACTION ITEM: jweiss will verify that we can add that much disk and  
will make the changes if possible

> VIII. Other Business

Ksplice wants us to use uptrack on the clusters.  We get rebootless  
kernel upgrades, they get exposure.   We have no strong objections to  
this, but it's very low priority because of all the other cluster  
issues at the moment.

ACTION ITEM: jdreed will respond to tabbott.

The next meeting will include discussion of migrating  
debathena.mit.edu to the VM currently known as athena10.mit.edu

home help back first fref pref prev next nref lref last post