QuickStart Instructions for the Auto QA Browser

• Go to the STAR Computing QA home page (from the STAR Home Page select "Computing" in the left panel, then select "QA" or go directly to the QA home page) and open the Auto QA browser by clicking on the "Offline Automated QA Browser" button at the top of the page.
• If the browser fails to open contact the QA Experts ASAP. If you cannot get to the Auto QA browser then you are S.O.L.
• Select data class = "Real Data Production" from the pull down menu in the upper, red area; the various query buttons appropriate for "Real Data Production" will reset after a few seconds.
• From these pull down menus make the following selections:
  • prodOptions = "any"
  • runID = "any" or a specific run number
  • job status (DST production) = "done"
  • job create Time (DST production) = "any" or "within last 14 days" or fewer if more jobs are retrieved than can be listed.
  • QA status = "Done, not analyzed" (QA job ran, may or may not be OK, job has not been marked as analyzed by shift crew)
  • QA job create Time = "within last 24 hours," depending on when the preceding QA shift was done.
  • dataset = "any"
• Click on the "Display datasets" button to fetch the jobs to look at. These will be listed in the upper panel. If you get too many or too few try adjusting the query options to be more or less restrictive. The "Real Data Production" jobs are top priority for the Offline QA shift crew. Note, at present this is only about 10% of the DST production jobs. This fraction may be adjusted by STAR management depending on the Offline QA shift crew's work load.
• Go thru a similar set of steps for the other Data Classes. Note that the query options differ for each class of data but should be clear once you have some practice.

• Note the "QA Status" information. Auto QA runs several macros for each DST production job. Each macro might or might not run successfully. Did all the QA macros run OK or were some not run or did some crash? If the bfcread_eventBranch macro ran OK, then you may still be able to check the job even if some of the other macros failed to run successfully. If many QA jobs crashed then notify 'starqa-hn' at once. You may ignore the various statements about "errors" and "warnings".
• In the right hand box select the "QA Details" button which will display links to the QA Shift histogram file(s) in the lower panel. The files to look at are listed immediately after the heading "QA histograms for shift crew." It is very helpful if your workstation is configured to automatically gunzip the files and display postscript directly. If not, the UNIX file location at RCF is given.
• First look at the file called EventQA_shift.ps.gz which is only 1 page and shows how many events were processed, how many had found primary vertices, how many events with a found primary vertex are in one of four multiplicity bins, plus some dE/dx and total charge deposition information. Depending on the trigger condition these four multiplicity bins may or may not have events.
• For min-bias runs there will be three files for low, middle and high multiplicity events, called EventQA_shiftLM.ps.gz, etc. For Monte Carlo Nightly Tests there is only one file called EventQA_shift.ps.gz. Select the file(s) to look at; if the LM, MM and HM files are present then you must check all three since some QA quantities such as track quality, for example, may vary with multiplicity.
• Review each plot in the file(s) comparing to examples (see Description of Offline QA Shift Histograms) and references (if available, these are listed with the StEventQAShift files). This is the main work to do while on shift. You should become familiar with the QA Shift plots and have some idea of what to expect before taking your shift. Unfortunately there is no easy way to provide a simple to use document describing the QA histograms which will always match page-by-page what you see in these EventQA_shift postscript files. The set of histograms depends on the particular set of Makers used in the reconstruction chain. However, you should be able to find an example in the Description of Offline QA Shift Histograms for any plot appearing in these ps files.
• After reviewing the QA plots scroll down the lower panel and select the "bfcread_eventBranch" button just after the heading "Scalars for QA shift crew." This will cause a new window to open with run summaries of scaler quantities.
• Review the scalers keeping in mind the multiplicity class of the events. The scalers include the min, max, mean and rms values for the number of hits, tracks and vertices for all the events in the job. Refer to the Description of Offline QA Shift Scalers document.
• Fill in the QA Shift Report for this job (see QA Shift Report Preparation and Instructions) and mark this job as having been analyzed by clicking the "Set QA Analyzed" button in the right hand box in the upper panel for this job. This will cause a box to appear in the lower panel for you to enter a password ("qashift").
• After completing all the listed jobs add whatever comments you think are useful and appropriate to the QA Shift Report for the "Real Data Production" jobs.
• Continue with the other Data Classes if available and if time permits. Note that the query fields will change for each data class.
• Continue to query for newly completed jobs throughout your 8 hour shift since DST-QA production is a continuous process (hopefully!).
• MOST IMPORTANT!!! If you suspect any problem with the detector(s), calibrations, reconstruction or production you must contact the appropriate expert(s). This is the basic reason for having the Auto QA system and these dedicated shifts. The experts may be contacted via either the QA Experts or Other Experts web pages. It would also be a good idea to consult with the Offline QA oversight team member who is on duty about the potential problem.
• Complete your QA Shift Report and submit it. The ASCII text version should be emailed to 'starqa-hn'. The web form (when available) can be submitted by clicking on the "Submit" button.
• Finally, you are done for the day; go get some rest!