[1357] in athena10

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

Re: Apt repository layout

daemon@ATHENA.MIT.EDU (Jonathan Reed)
Sat Mar 7 14:01:17 2009

Cc: Evan Broder <broder@mit.edu>, Tim Abbott <tabbott@mit.edu>,
   debathena@mit.edu
Message-Id: <0902A9EE-2996-4ACC-8F74-086B7924B575@mit.edu>
From: Jonathan Reed <jdreed@MIT.EDU>
To: Greg Price <price@mit.edu>
In-Reply-To: <20090307174449.GA3488@cross-cap.csail.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: Sat, 7 Mar 2009 13:51:37 -0500


On Mar 7, 2009, at 12:45 PM, Greg Price wrote:

> SIPB office heads should use -proposed and run -auto-update; in  
> fact, they
> should probably run -managed (or whatever name it gets).  That will
> provide some testing.  If the staff involved with Debathena also run
> -auto-update and use -proposed on their desktops, that probably  
> suffices.

I think this is a good idea.

This essentially mirrors the functionality we have now with Athena 9  
"alpha", "beta", and "public" clusters.  Since we haven't had a major  
release in 4 years, the existence of the "alpha" cluster may have  
fallen off people's radar.  Basically, machines in the "alpha" cluster  
get a bleeding release when we do a major version update.   
Historically that was 1 OLC machine, 1 SIPB machine, some developer  
machines and a few other random machines.  When we're not in the  
process of doing a major release, alpha machines behave like beta  
ones.  The "beta" cluster (aka "dev cell machines") gets updates  
before they go out in the field to ensure that things aren't broken.   
Historically, this was another OLC machine, another SIPB machine, and  
desktop machines of Athena folk in I/S (&T) (me, alexp, wdc, boojum,  
ops folks, etc).

So, if we go this route, I would in fact propose that we add  
clusterinfo data to the alpha and beta linux clusters, which would  
cause them to take -development and -proposed, respectively.   
Obviously, we'd want to notify maintainers of existing machines prior  
to this change, and allow them to move into different clusters prior  
to installing Debathena.   But I think using our existing pools of  
machines (and updating membership of those pools as needed) is the  
best way to ensure that new features get a lot of testing in the field.

-Jon

(If you're interested in seeing cluster membership, "qy -s gmcm \*  
alpha-linux" and "qy -s gmcm \* beta-linux" are the relevant moira  
queries.)

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