[6622] in Release_7.7_team
Minutes of Release-Team meeting, tomorrow, 2pm, N42-203
daemon@ATHENA.MIT.EDU (Jonathan Reed)
Sat Feb 13 18:54:11 2010
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0 (Apple Message framework v1077)
From: Jonathan Reed <jdreed@MIT.EDU>
In-Reply-To: <23C3794D-C335-49D4-A9FB-67D35CDFCCE4@mit.edu>
Date: Sat, 13 Feb 2010 18:54:03 -0500
Message-Id: <0D208883-AED9-4347-93F5-DA206C2658A2@MIT.EDU>
To: release-team@MIT.EDU
Content-Transfer-Encoding: 8bit
Attending: rbasch, othomas, mmanley, amb, jdreed, geofft, alexp, aurora, broder
> 1) Begin transition of CUPS.MIT.EDU hostname to ops
The change in was made in moira on Friday and is now active:
cups.mit.edu is an alias for CLUSTER-PRINTERS.mit.edu.
CLUSTER-PRINTERS.mit.edu has address 18.9.63.33
SIPB has contacted users who requested private queues to be added to CUPS.MIT.EDU to let them know of the change.
> 2) Review proposed auto-updater changes and prioritize them
ACTION ITEM (for all team members):
Clarification is required on the following tickets. Trac updates and/or closing/re-milestoning are required:
-#412: Isn't this already done? I thought we already changed the desync interval and did this?
-#217: Does this still apply in the new login chroot world? And will it still apply depending on how we address #309?
-#464: There seems to be disagreement on whether this is a bug or not.
Remaining tickets, In rough order of maximize(priority, simplicity):
#407 is by far the easiest to fix, which is to simply have the auto-updater run "dpkg --configure -a" on each invocation (which is a no-op if there's nothing to do).
#516 if auto-update fails, we should syslog with a known unique string, and coordinate with ops to get this information into the gazette
#406 is similar to #407 (the auto-updater does nothing if there's nothing to install but some packages are unconfigured. Possibly this is not a problem if we fix #407, since it will always try to reconfigure packages.
#469 is the ticket to either get desync interval from a URL, or to fetch a whitelist of "upgrade now!" packages from a URL. This will help us deploy quick fixes.
#309 primarily affects not-cluster machines. Greg's suggestion of desync -t seems it would make a simple fix.
#483 is a two part ticket. The fix for --cluster is easy and known. The fix for workstation blocks on a nice way to notify the user. Perhaps we should implement the first fix sooner rather than later.
#408 This is lower priority, since it really only bit us on Intrepid->Jaunty
> 3) Karmic update
cluster-login-config builds, but event.d moved so our tty login changes may not work. Testing is required. It's possible -cluster still barfs on nmh-config, which is blocking on collaboration between Mitch and Geoff. jdreed will do a test install on Tuesday (others should too).
The new gdm puts a dmrc file in /var/cache/something, which may cause problems for repeat logins on the same machine.
ACTION ITEM: rbasch will open a ticket describing this situation in more detail
Should we punt to Lucid? No, this work still needs to get done.
We still believe we're on track for an Add Date deployment.
-thirdparty is uninstallable. There is a ticket (#515) assigned to jdreed.
QUERY: Do we want to (manually) install karmic on a few select machines once it's installable and usable? Probably. We don't need anywhere near the number of beta machines we had for Jaunty. It's possibly a row of W20 and next5/6 will suffice.
> 4) demeter
The pros and cons of moving away from scripts are debated. Scripts has instant failover capability, but ops has the ability quickly restore a dead VM. Demeter gets us the ability to: use NFS mounts for the livecd instead of mod_rewrite tricks, a static IP for the installer (though Scripts is willing to special-case that). There are no strong feelings either way, but if we lose demeter, we're unlikely to get it back.
ACTION ITEM: jdreed will replicate debathena.mit.edu on demeter and submit it for proofing at a later release-team.
