Minutes SLAC-ALTAS TDAQ Meeting 05-Oct-2006
-------------------------------------------

Present: Sarah, Amedeo (phone), Stephen, Rainer


Sarah:

Looked at the ability to access files from dbproxy.  Need to talk to
Radevan (sp?) about file transfer.  Same person to consult about the
'use' statements (i.e. the non-atomic queries).

No way to do the test with the partitions as they are now and throw in
a different CORAL version, because mismatch too large.  Will talk to
the CORAL people to have them produce a patch that can be used with
the partition.


Amedeo:

Wants to make two changes: Change the dependency in the way the
partition is created, such that all proxies are not started at the
same time.


Sarah/Amedeo:

Since there is the need is to start proxy is the proper order in order
to build up the tree, want to build the correct dependencies into the
partition.

Once the proxy is alive it sends a PMG (process manager) call to Run
Control to announce it is ready.

First try did not work, because Run Control time out was not set
(i.e. zero).  Giovanna suggest to set it to 10s.  Does not meen that
RC will wait for 10 s but that it will time out then if it didn't
receive the PMG acknowledge.

Most developers include PMG sync as a command line option to the
application.  But we don't need that, because dbproxy already knows
whether or not it is in an IPC environment.

ERS was not working in tests with Serge.  Sarah was not able to
reproduce that on lxplus or anywhere else.  Will talk to him to try
and run on his partition.

Talked to Sergey Kolos, but he cannot debug this without seeing it
fail.  Sarah working on that.

[Sarah reiterates apologies for misspelling Steffen's name.]

MST: Abandoned TDAQ 1.4 for now, since could not get this to work. Now
trying with TDAQ 1.6.

Discussion on how to proceed with SLAC farm tests when MST is
officially declared over (second week of Oct).  Feel that SLAC tests
are worthwhile anyway.  Given the experiences with 1.4 want to skip
ahead to TDAQ 1.6 and give that a shot.  May abondon the HLT image
that was going to be prepared for MST.

Settling on the ERS business. Will use a command line option to make
debug output, including dumps, run time switchable.  This will replace
the #ifdef DEBUG.  Eventually can still compile out the debug switches
if performace is critical.


Stephen:

Talking to Jiri (sp?) about adding tests to the AtlasHLT project.  Not
part of the release (for some reason).

There is some backpressure from the offline against including more
code into the nightlies, even though this seems wrong, and it should
be desirable to build as much (relevant) code as possible, including
online.

Zach needs 12.0.3 at SLAC and that's there now.  He needs a debug
version to see symbols in profiling (the shower parameterization
code).

He also asked for a nightly build to be installed.  This comes from a
different mirror.  Will check.

Intructions for installating ATLAS releases that are in the workbook
or main Wiki should work.