Minutes of NAF User Committee meeting from 08.07.2009 
-----------------------------------------------------

Present/Phone: Wolfgang Ehrenfeld (ATLAS), Andreas Gellrich (NAF),
Andreas Haupt (IT), Kai Leffhalm (NAF), Niels Meyer (ILC), Hartmut
Stadie (CMS), Jan Erik Sundermann (ATLAS)


1. News from the chair:

Last week the 'Detector Understanding first First Data' workshop took
place at Hamburg. The tutorials for ATLAS and CMS were running very
smoothly on the NAF. Thanks to all the help from the NAF admins to
achieve this.


2. Action items:

Only items with updates, see action item list for complete list of
items.

 - ATLAS cmt compile problem

During the detector understanding workshop the ATLAS software
installation was mirrored to the local disc of all ATLAS work group
servers. This was working very good and scaled very well for many
people working and compiling at the same time. ATLAS will discuss with
the NAF admins what the prospects are to have part of the ATLAS
software permanently installed on the local disc of each work group
server. CMS suggested to increase the AFS cache to have a full release
in the cache. 

 - NAF SL5 migration

There are no new updates on SL5 software compatibility from ATLAS and
CMS on testing. ILC testing is ongoing and some problems were
discovered and fixed.

CERN will migrate their lxplus and lxbatch service sometime this
summer, but details are not available.

After some discussion the NUC agreed that 90% of the batch resources
will be switched to SL5 shortly after the next NUC meeting in
August. If contrary to expectations  major concern from users show up the
migration plan might be altered at the next NUC meeting.

The NUC chair will inform the users and ask again to test their
software under SL5.

The NAF admins will inform the users when the default OS for batch job
submission will be changed from SL4 to SL5. This will happen when
around 50% of the resources are migrated.

Every experiment has by now a dedicated SL5 work group server.


 - batch monitoring plots: waiting time by experiment

Some improvements have been added to the batch monitoring jobs,
e. g. waiting time is split up by experiments. See last page of status
report. The list of suggestions by Hartmut was longer and he will
discuss the details with Kai.


3. Status report:

Status report was given by Kai. See the agenda for the report. Below
a few highlights from the discussion are listed:

Currently the Lustre server has some problems. A reboot usually
helps. Problem is under investigation by the NAF admins and
sometimes the restart of the service is delayed to do further
investigations.

Hartmut pointed out that important user data is stored on Lustre. The
disruption of the Lustre service has the same impact as a disruption
of the AFS service. A quick restart of the Lustre service is encouraged.

Wolfgang proposed that the NAF admins will inform at least the NUC or
the support teams when there is repeated loose of a service. Not all
users will complain to the NAF helpdesk for operational problems but
will complain to the experiments support.


4. AOB:

 - SL4 and SL5 have the same sysname. A different one for SL5 is
   essential for proper software installation for SL4 and SL5. The NAF
   admins will report back on this issue well before the next meeting.

 - Hartmut said that it is difficult to keep the users up to date
   about new software tools at the NAF and suggested to have a news
   letter on a regular basis to explain these. An alternative is the
   use of motd to inform users. The NAF admins pointed to the news
   section on the NAF web, which covers some aspects of user
   information.

   The NAF admins should reevaluate the user information and optimise
   if necessary.

 - Wolfgang suggested to enable the automatic reply feature of the
   ticketing system (naf-helpdesk@desy.de and naf@desy.de). This will
   give the user/admin a ticket number to reply to in case of no
   response. The NAF admin will investigate if this is possible.