Actions arising from the Quattor Workshop
-----------------------------------------

(F. Minafra): Francesco suggested having a quattor tutorial in
conjunction with the Tier 2 meeting at CERN 12-16 June.  German wants
to verify that there is enough interest to justify the effort to do
this.  The general concensus was that there would be enough
students.  Past quattor tutorials have scared people off because of
the complexity and because there were not sufficiently focused on a
particular problems (e.g. LCG installation).  It was felt that such a
tutorial should be focused on a single, simple solution and give
concrete, real-world examples.  Francesco agreed to coordinate the
planning for this among those interesting in contributing. 

(C. Loomis, S. Childs, S. De Weirdt): Work to converge on the versions
of templates/build files used for the QWG LCG/EGEE configuration.  

(C. Loomis): Ensure that those collaborating on the QWG configuration
have write access to the QWG trac and to the QWG repository. 

(C. Loomis): Look into modifying the quattor build tools to generate
component templates compatible with namespaces, rpm distribution of
components, and profile distribution of components.  

(G. Cancio): Determine the current level of GridIce and LEMON
integration and send to the QWG a summary.

(C. Loomis): Put the current Quattor-based LCG Installation Guide onto
the QWG wiki.

(S. Childs): Update the Quattor-based LCG Installation Guide.

(C. Loomis): Feed back current GRIF configuration/organization into
the QWG repository.  (Build files and structure.)

(R. Starink): Prepare a new version of AII for the next Quattor
release.  This should include the --rescue option, fixes for the
outstanding AII bugs, and fixes for the recent bugs in the kickstart
template.  If possible include changes to help with Mattias' problems
with diskless systems. 

(G. Cancio): Send information to the list on the current version of
LEMON and any available documentation. 

(C. Loomis): Incorporate following changes to the current
implementation of the pan compiler: 1) 'unique' template capability,
2) include statement which takes a DML block as argument, 3)
specification of default values in type and record statements. 

(C. Loomis): Rewrite the pan compiler to improve the performance of
the compiler.  (Better memory management and multi-threaded
compilation.) 

(S. De Weirdt): Look into applying German's trick with fake object
templates to avoid long recompilations of package lists for identical
machines (principally worker nodes).  

(All): Send feedback to German Cancio on which CERN-specific
components others would like to use.  This will help to concentrate
effort to generalize those components. 

(G. Cancio): All agreed that there needs to be a better way of finding
out what components exist, their state, and stable version.  German
will make a first stab at automatic generation of this information
from the CVS repository.  Others will then comment on what needs to be
improved, added, or changed. 

(M.E. Poleggi): After the namespacing is available in CDB, pull
together a new release of the base quattor services.  The approximate
timescale for new versions of services to be ready is two weeks.

(G.Cancio): Circulate a proposal to the list on how to manage the
lifecycle of the bugs and tasks for the group.  German will discuss
this with the ETICS project to see what can be done in common. 

(C. Loomis): Update the standard pan and quattor templates to use
namespaces.  These should be in the pan and quattor namespaces,
respectively. 


Further Quattor Workshops
-------------------------

The general concensus was that the workshop was useful and that we
should aim to have one every six months.  We will aim to have the next
one in the first part of October 2006.  The location is to be
determined by concensus and who wants to volunteer to host it. 

This workshop lasted 3 days.  It was felt that this was appropriate so
that people could get to know each other and to understand the current
state of the quattor services and configurations.  For future
workshops, probably two days of meetings spread over three days is
probably sufficient.  (Allowing people to travel the morning of the
first day and leave the afternoon of the last day.)

The workshop agenda left a large amount of time for discussion.  All
felt it was extemely important that future workshops maintain this.

A possibility for future workshops is to allow 'breakout' sessions on
specific topics which might not concern everyone at the workshop.
One example of this to have the CDB and SCDB discussions separately. 

Another session of general interest would be a 'tricks and tips'
session where everyone can share their experiences with making cluster
management easier or more efficient.