Minutes of the Data Quality Meeting of Friday 22 August 2008

• Indico agenda page

Procedure

Login problems prevented PK from uploading transparencies. An updated version will be uploaded later.

There was a lot of discussion about the DQ procedure. The bottom line is that many things need to be precised.

Main issues were:

• We need a *clearer procedure on how to deal with problems requiring software or database changes.
  • Once a problem is identified, an expert is asked to provide an updated Brunel or databse (who tags?) and a test job is submitted. Philippe offered to set up an infrastructure to do so.
• The DQ shifter should be in charge of the day-to-day job: look at reconstruction histograms, make sure all is OK, follow up that calibration is done, submit and analyses control jobs.
  • The DQ shifter would have close contact with the offline shifters and work in the offline control room.
• A DQ expert on piquet should have the long term overview of what is hoing on. He would run the daily DQ meetings and be responsible for the decisions.
• The daily meetings should not be linked to run meetings as previously discussed. The DQ meetings would rather be in the afternoon, after the offline computing meetings, in the offline control room.
  • The DQ shifter and experts would either attend the run meeting, or be informed about the problems discussed. They would report at the DQ meeting for further discussion.
• It is agreed that anyone should be allowed to report problems in the Problems Database. LHCb shifters should be encouraged to do so. Subdetector experts can also report problems, but it shuld be assumed that only experts would do so, as there would be a risk if having only the people reporting problems who would then be assigned it for solving.

Calibration stream

There was a short re-iteration of the dicussion. Having a dedicated calibration stream implies work and will not be implemented in 2008. We have to revise the requirements for 2009. It is understood that we need raw data as calibration jobs usually require re-running the reconstruction. This can be done on the plus farm or using castor data. The procedure needs to be defined.

OMAlib

Does not depend on Online. Patrick continuing investigations.

Savannah

We should encourage every LHCb member to have a Savannah account. All LHCb and DQ shifters must be Savannah users.

E-log

We need to set up a DQ e-log.

-- PatrickKoppenburg - 22 Aug 2008