AF
meeting 29 May 2008
Attendees
Gabriele Cosmo (chair), Stefan
Roiser, Marco Cattaneo, Predrag
Buncic, David Quarrie, Fons Rademakers,
Witold Pokorski, Rene Brun, Oliver Keeble
Agenda
- Line
management issues
- Action
items
- Status
of releases
- Platform
and external libraries issues
- Experiments
feedback
- R&D
project status
- Topics
for AA meetings
- AoB
version 1 29/05/08 @ 16:00)
Line
management issues (Pere)
- Contract
extensions of two months have been granted to Leo Franco
(Virtualization R&D project) and three months to Tatiana
Nikitina (Geant4).
Action
items
- Cross
compilation of ROOT (32 vs. 64 bit) (Fons)
- Fixed,
action can be closed.
- Legacy
support for ROOT 4.x files by POOL. Confirmation from ATLAS (David)
- Support
can be dropped, action can be closed.
Status
of releases
- Preparation
of LCG_55 (de-Sealed version)
- The
request for a "de-Sealed" version of the LCG software is coming from
LHCb; Marco reports that changes are being made in Gaudi and it is in
plan to release Gaudi towards the end of next week.
- The
idea is to freeze the application area stack; LCG_55 will use
Root-5.18/00d (including few fixes for BNL)
- Planning
release LCG_54g following request by ATLAS; the release will include
Root-5.18/00d , new Coral and FastJet- 2.3.3.
Platform
and External Libraries issues
- Supported
platform for PVSS 3.8 [1]
- PVSS
for SLC5 is of course wished. Marco made notice that SLC5 comes
equipped with gcc-4.1, previously it was though to move the LCG
software directly to gcc-4.3 but pending the new release gcc-4.3.1 to
be announced; issues may come from Castor/Oracle clients; a possibility
is to keep building on 4.1 and move to 4.3.X later (SLC6 ?).
- Oliver
reported that EGEE/TMB (Technical Management Board) has decided for
assuring Python backward compatibility; version 2.4 of Python is by
default in SLC5 but bugs reported for 2.5 will be considered. LCG
software is based on Python 2.5 since quite some time.
- Pedrag
notice that few packages (among which Python) are the responsible for
real binding to the operative-system; experiments use to copy over all
the rest.
- It
was decided for the moment to go for gcc-4.1builds for the nightlies
- Migration
to SLC5.
- Stefan
reported that a pilot SL5 installation will be done soon on one of the
available machines.
- Oliver
reported that following
the recent Management Board discussion, FIO intend
to have SLC5 ready and deployed by the end of September; the middleware
will also get ready for this, no code changes are expected for this
release except for few bug fixes; the migration is expected to happen
only after a stable version has been tested on SL5 first.
- Marco
makes notice that for the experiments the priority for the next 18
months is certainly to not change the OS but still keep SLC4. Oliver
makes notice that Red-Hat support for back-ported updates on RHEL4 is
now ending.
- Stefan
observed that Linux certification meetings are held every two months
(Stefan representing SFT in the meetings) and that would
be good if the LCG software is synchronised with the certification
process.
- Windows
VC++ 9.0
- Stefan
reported that he is now starting to set up the night-lies; some issues
have come up with CMT producing wrong makefiles, the issue is still to
be investigated. Marco made notice that by using the free version of
VC++ some executables which are used by CMT used to be part only of the
licenced-version, to be verified if this applies to VC 9.0 as well.
- MacOSX
10.5 (Leopard)
- Builds
on Tiger are now dropped; building only on Leopard, and so far things
are going well. There's still one machine with Tiger left (macsft02),
this may be kept for some time still before upgrading it to Leopard as
well.
- David
raised the question if to migrate the build of libraries to .so for
shared ? Problems
affecting Tiger for dylibs are now resolved in Leopard and having .so
or .dylib makes no longer difference. It was noticed that CMT
explicitely require the .dylib suffix; it was agreed to proceed asking
for changing the configuration in CMT and later on also Gaudi will
move; the modification will take place in patch to both LCG_54 and
LCG_55.
- Changes
in the externals
- A request has come from LHCb for the installation of the libSVM package.
- No
changes, except for the addition of libSVM which is now installed.
- POOL:
- Witek
reported about the effort currently going for reducing the complexity
of the package; a meeting with representatives from the experiments has
been held to know about their use of POOL. It was sorted out that no
much cleanup can be done since almost all the modules are somehow used
by the software in the experiments. Witek made notice that having POOL
as a common layer for accessing the different storage-services is not
really exploited
by the experiments, some use the Root storage-service of Pool, others
only the relational part ... Therefore, only few modules (which are not
needed) will be removed
- ATLAS
and CMS will remove the dependency on the data-service module in the
next 1-2 years.
- Would
like to start a new CVS for POOL for cleanup; it was raised the
question if it should be considered to move to SVN directly instead;
David noticed that it would be good if all the LCG software could move
to SVN together instead of having independent migrations by each
project. It was therefore decided for now to start a new CVS module in
the next month or two and eventually move to SVN later on as part of a
common LCG strategy, if decided to do so.
- It
is intended to review the contribution to the POOL project from the
experiments; contribution which has been reduced substantially in the
past months; it is necessary to reassign the responsibilities of
experiments to the various projects (for example, the file-catalog is
assigned to CMS but CMS is not using it, while it is used by ATLAS ...)
for better managing developments and maintenance.
- Schema
evolution proposal:
- Witek
reported that there is no problem with the Root proposal from the POOL
side; it should be discussed how the experiments intend to use the
schema evolution, from the POOL side it will be transparent.
- Marco
reported that a concrete example has been provided by LHCb.
- René
reported that M.Novak from ATLAS has expressed positive feedback for
the proposal.
- Persistency
projecct:
- Progress
is being made on the Coral side.
Experiments
feedback
- CMS
- LHCb
- Nothing
particular to report. Marco asked about progress on the requested fix
in Root for the HepMC-2 plugin; René reported
that the fix is coming and will be available in version 5.18/00d; for
the schema evolution a special streamer has been developed.
- ATLAS
- Nothing
additional to report.
- Alice
R&D
project status
- Virtualization
project (Pedrag)
- The
new 0.5 release is under preparation; the release includes small fixes
and extra packages on request from LHCb, addressing SLC4/5
compatibility libraries; it was noticed that some of the externals are
still linked to SLC3 libraries instead of SLC4, this is being fixed.
The new release also added an utility for ATLAS to import the entire
directory tree in the cache (same as keeping it locally in the virtual
machine). It has been added support to specify how to configure
secondary proxies; proxies can be auto-configured depending on the
location, feature useful particularly for laptop users.
- It
has been added support for CMS, and the image is now available. All
four experiments are now covered. Detailed testing is required from
experts in the experiments, also for what concern a performance
evaluation.
- From server side, the infrastructure can scale as needed.
- The
new hardware has not been ordered yet; using simple servers recuperated
from Alice.
- The
URD document still to be completed.
Topics
for future AA meetings
- 4
June :
- ROOT
schema evolution proposal (Lucasz)
- 25
June :
- ATLAS
software validation (David)
AoB
- Next
AF meeting in two weeks.
[1] Question coming from ETM via Markus
Frank and IT-CO:
ETM
should now define the supported platforms for PVSS 3.8.
They
are thinking about moving from RHEL4 to RHEL5.1 and no longer support
RHEL4.
What
is CERNs position about this ?