[276] in Athena Release Announcements

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

Athena automatic update failure; intervention required

daemon@ATHENA.MIT.EDU (Jonathan Reed)
Wed Feb 23 12:54:09 2011

From: Jonathan Reed <jdreed@mit.edu>
Content-Type: text/plain; charset=us-ascii
Date: Wed, 23 Feb 2011 12:53:35 -0500
Message-Id: <C04ED7CF-F7A8-425F-A2A5-38CBFD1F99FE@mit.edu>
To: release-announce@mit.edu
Mime-Version: 1.0 (Apple Message framework v1082)
Content-Transfer-Encoding: 8bit
Reply-To: "release-team@MIT.EDU Release Team" <release-team@mit.edu>
Errors-To: release-announce-bounces@mit.edu

Hello, 

This message only applies to users running the "debathena-workstation" or "debathena-cluster" metapackages.  If you're unsure which metapackage you're using, run the following command on your workstation:
     machtype -L

At the end of January, most Debathena workstations across campus experienced a failure involving the automatic updater (technical information is included at the bottom of this e-mail).  Unfortunately, this failure is not recoverable without human intervention.  Workstations should still be fully functional, but will not take additional updates until the problem is rectified.  The IS&T Cluster Services team has been fixing public, IS&T-owned workstations throughout campus, however private workstation owners will need to take the following steps to rectify their workstations.  

This procedure should take anywhere from 1 to 5 minutes.

1) Become root (either with the "su" command, or "sudo -i")

NOTE: debathena-cluster users should instead reboot their worksation into recovery mode and obtain a root shell, as described at http://kb.mit.edu/confluence/x/tgCt

2) Run the following commands:
    apt-get -y -f install at
    dpkg --configure -a
3) Reboot your workstation

If you have any questions or concerns, please contact release-team@mit.edu.

We apologize for the inconvenience.

-Jonathan Reed
for the IS&T Athena Release Team.

Technical information:
The automatic updater uses a desynchronization process so that every workstation on campus does not update simultaneously (which would drastically increase load on the network and servers).  In the past, this was accomplished through a cron job which called sleep(3) with a random value.  Unfortunately, modern operating systems discourage sleeping during cron jobs, as this confuses the system's understanding of which users are logged in at the time.  The Athena Release Team therefore decided to use the "at" task scheduler, which provided the added benefit of allowing private workstations to enter power saving mode more efficiently, thus contributing to the "Greening MIT" effort.  Unfortunately, an update to the "at" package caused the scheduler to kill all running tasks, including the update, leaving the "at" package in a broken state and unable to resume scheduled updates.  We believe this failure to be a bug on the part of the Ubuntu maintainers.  To avoid future occurrence!
 s, we are releasing a new version of the auto-updater which does not use the "at" scheduler, yet still allows workstations to manage their power usage efficiently.


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