Minutes SLAC-ALTAS TDAQ Meeting 30-Nov-2006
-------------------------------------------

Present: Stephen, Sarah, Steffen, Matthias, Rainer


Stephen:

Talked to Frank Winklmeier.  Running different slices with Athena,
also going to do it with AthentaMT and AthenaPT.  Scripts need to be
adapted.  Has its own script language.  Stephen has been using
regression test from BaBar, which is much simpler.  Discussed which
version Frank wants to use.

Comparing log files output in the nightly.  The log files have certain
lines that are meant for the regression tests.  With multi-threaded
they come out in different orders. (Event though there is supposedly
only one thread executing the code.) Have been grep'ing for the lines
and sorting them.

Release 12.0.3 is where the tests were developed.  12.0.4 and nightly
failed until the reference was updated.

An obvious other thing to do is testing histograms.  Discussion about
testing timing as well.  Could do delta's between 'last night' and
'tonight'.


Sarah:

Tau slice was added to the nightly yesterday.  Job options files are
available now in the trigger release area.  Will be a part of the
LST-2 release.

LST: Partition maker has not icorprorated dbproxy properly.  Have a
partition that has two layers of proxies.  This is new from what Serge
had last week.  Partition Maker creates the xml file, xml file is used
by PMG.  The proxy is not started by the PMG agent (process manager).
Needs to be sorted out.  All interactive tests work perfectly when
proxies are started by scripts.

Lorenzo has made some progress with the partition structure, running
750 nodes.  Stressing the Oracle DB in the way the detectors will.

The HLT image is not yet working for anyone.  Also tried to bring it
over to SLAC and ran into technical problems with that.

Image about two months behind schedule.

Will run LST till Dec 10.

For the proxy can always create a large partition by hand in order to
get tests done.  Would make a lot more sense to have partition maker
right, including dependencies.  But have this in.  Rodrigo is working
on partition maker has a newer version but that needs new code than is
in the LST release.

Mysql access to geometry still not working.  Needs switch between
atlasdb and atlascool.

Suggestion to test the caching by restarting a configuration cycle
without bringing everything back up again, i.e. in a warm start.


Amedeo:

Implemented the expiration of the oldes entries in the cache.  By far
the most complicated algorithm in dbproxy just now.

Expiration is by first use, not last use.  It's not the ideal thing in
terms of maximizing cache hits.  But the advantage of this is that the
pool has only one gap that's moving, not many holes.