meeting 3rd February 2011
Clemencic, Benedikt Hegner, Fabrizio Furano, David
Rousseau, John Harvey, Peter Hristov, Liz
Sexton-Kennedy, Gabriele Cosmo, Pere Mato, Rene Brun, Stephen
Gowdy, Fons Rademakers, Predrag Buncic, Vincenzo Innocente,
Andrea Valassi, Marco Cattaneo
Line Management Issues (John)
Report from the LIM meeting (Benedikt)
News from projects
1 4/2/2011 @ 17:00)
Stephen Gowdy was introduced by John Harvey. He is a co-convener of the
CMS core software with responsibility in the release management.
Management Issues (John)
- New Technical Student, Alex Hodgkins, will be working with
Benedikt on the LCG nightlies.
- The NA61 software upgrade proposal review will take place on
February 10th. There will be an open session from 14:00 where everybody
- The SFT internal deadline for the Summer Student's requests is
next Thursday. A total of about 7 students are expected for the
group. Therefore a couple of proposals per project would be adequate.
what python modules exist in the external area (Benedikt)
- In fact the lack of documentation for the external
packages is more endemic and not limited to the python modules.
Documentation is also incomplete for many other packages. This is work
in progress. 'Story' entered in the 'To-do' list.
of the status of the '_cov' platform in
the nightlies (Benedikt)
- Benedikt reported that the coverage platform is
working now. Done.
the 5 to 10 most important requests/requirements for each of the
projects (Simulation, ROOT, Persistency, SPI, CernVM, Multi-core) (ALL
- See later point in the agenda.
MacOSX and Windows on the nightlies need a fix ASAP (Benedikt)
Planning Preparation (Pere)
- Pere reported on the selection of the JIRA tool to support
project planning. SPI is currently pursuing a production installation
with definitive license, which requires that the server is visible to
the outside world.
- Pere informed that he got the input from CMS yesterday by
participating to their project leader's meeting. The input needs to be
phrased correctly to be put in the preliminary plan but is a very good
starting point. This method of interaction with the experiments has
been very useful and it would be nice if it can be made also for the
rest of the experiments. It was left to the experiments to decide what
is the best for them.
- As soon as we have more input we will circulate it for comments
and later discussion in the AF.
Report from the LIM meeting
- Moved the meeting slot to Tuesday 14:00. Using 32-S-C22
instead of the restaurant 1 coffee area.
- ATLAS starting on clang/llvm. slc6 beta expected
around early March. Agreed to set up nightly slots for both platforms
- Timescales and priorities to be discussed in the AF. See later
- ATLAS built and deployed SW on top of LCG_59b.
still in the process of building on top of LCG_60.
- ATLAS adapting their
nightlies to LCG60 now. They want to have additions to it (e.g. newer
frontier, fixed ROOT pcre issue) before moving to this series.
- Available new versions of
lcov, neurobayes, valgrind,frontier_client to be included into the
- Discussion on action items from last AF
- Coverage testing and
BuildMover tasks considered done.
- Work on the python package
documentation revealed further gaps in the documentation.
- Project planning
- Persistency project will start integrating
major changes into preview tags soon.
Understanding the delays of last LCG releases
A number of actions has been identified in order to mitigate the causes
of the delay (about 2 weeks) in producing the last two complete LCG
- Add additional tests in the nightlies to check that all externals
(including the ones not directly used by any AA product) are available
for all required platforms for which the release is targeted. This is
particularly necessary for windows.
- Make sure that the same procedure is used to build ROOT in the
nightlies and when is released in AFS. In particular making use that
all the configuration options are identical.
- Circulate in advance all the changes introduced in the
to the experiment's representative to avoid late discovery of
Platform Support Discussion
- Windows (vc9)
- LHCb is aware that it takes a lot of time and effort to support
this platform. The benefits in the past have been to identify problems
that were invisible on SL/gcc. It is not clear whether this is still
the case and the overall balance is still positive. It was agreed to
continue supporting Windows for the rest of the year (on best effort)
and let LHCb proof during this time that there is a clear benefit. The
best effort level is acceptable for LHCb (mitigated urgency in case the
- The clang/llvm is coming into shape but still requires some
work. We can take either a released version
or the trunk. Fons suggested to move to the trunk as way to help
developers providing feedback. No conclusion on this point.
- Fons informed that icc version 12 is buggy. ROOT does not
compile (internal compiler error) and they gave up. On view of this we
should stay with version 11.
- SLC6 (gcc-4.4.x)
- When the beta version appears in early March we should add the
platform in the nightlies. Although the experiments are very likely not
going to adopt it for very long time (end 2012?). Marco C. said that
the important dependency is in the compiler version and we should
continue decoupling the compiler from the OS as we have been doing in
- Understanding the benefits and difficulties of a new version of
the compiler (gcc-max) and the new kernel should be at the R&D
level of effort (from Vincenzo).
- The interest of SL6 for CMS is the fine-grain control for
limits (i.e. network
bandwidth per program). This is probably a feature of the kernel.
News from Projects
- Proposed a presentation to the AF of the ongoing work in
restructuring of the XROOT client code and improvements for WAN
transfers from ROOT. Dirk will be contacted to schedule it ==> Action Pere.
- Little feedback to
the request for input. Vincenzo mentioned the possible interest in the
lxplus interactive service to make use of all the cores
- The 2.2.1 update was released. There have been some problem
with the update of the AFS kernel module that was preventing the
- Progress in the service implementation in IT. The proposed
storage solution is based on AFS and cannot provide the snapshot
capability, which is essential for the service. Nor clear yet the what
(virtual) machines will be provided to the experiments to populate the
software repositories. The advantage of using CernVM is that users can
see the exactly the same
platform/environment while running their tests after installation. The
experiments should say what they want. For LHCb the CernVM-FS is a
service. If there are problems they should complain to
MB (WLCG). The date mentioned was beginning of March.
- Started testing the
replication procedures between sites (RAL, BNL). No feedback yet
- Planning to migrate to a new namespace
(/cvmfs/cern.ch/<experiment>). Currently testing the
implications. ATLAS may have a problem in relocation.
- Other ongoing tasks
- make the client aware of
the multiple replicas
- update the desktop environment
contextualization mechanism requested by ATLAS.
- Development version of Geant4 released last Monday. Working on
collecting all fixes introduced so far and planning a public
patch 9.4.p01 in the coming weeks.
- A nasty memory leak found in ALICE was
fixed in both the 5.28 and 5.27 patch branches.
- Planning to use LCG_59b
during the data-taking period this summer.
- Preparing a new release base on Geant4
9.4, ROOT 5.28 to be used in September.
- They are interested when the parallel
buffer merge will be available from ROOT. Fons said that writing from
multiple threads to
a single file would not be possible.
- David asked on the status of the action from the Savannah
feedback document they presented some time ago. ==>
- Testing the new ROOT in their
- They have finally moved to 64 bit for online/offline.
- Started to campaign to
reduce the size of the data output. The ROOT feature (Double_32) is not
exactly working as they
- Liz asked if the Geant4 architecture review
document could be given to the experiments. Probably it could be
distributed with a big warning indicating the actual completion status.
- Finalizing the LCG_60 based release.
- The platforms they will support are SL5/gcc-4.3 (32/64),
and Windows. The 32 bit for SL5 is still needed due to an
issue with the simulation. There is also some reluctance to declare
discontinue the 32 bit by people running with old laptops.
- The online
has moved to SL5(64) with ROOT 5.27.
currently in the offline ROOT 5.28
- CMake migration very successful. Very negligible