QUICK RECIPE TO CHECK RUNS
||Setting a working environment
In the whole guidance, it is assumed that you're interested in looking at runs which have been processed in Event Reconstruction (ER). This is the second step of the processing -- after the Prompt Calibration (PC) -- which is made with more statistics than the first step. Therefore, it is the suitable processing to test the runs QA.
- The package RqmRootTreeTools is release-independent so any recent release should be fine. Yf you're using a too old release, you may have trouble in processing the root QA files. To run properly, the ROOT version of your release must not be older than the one which was used by OPR to produce the files. So, if you have such problems, compare the releases and update yours if needed.
- Checkout the package in your test release. You can get the head version with addpkg -h RqmRootTreeTools but it is not guaranteed to be working fine. It is safer to find the last official tag (VXX-YY-ZZ), either in the Data Quality Hypernews or directly in the CVS repository. The current manager of this package is Frank Winklmeier -- his term ended September 1rst but nobody was found so far to replace him.
- Compile it:
- Go to your workdir. Copy in this directory the file ER_Default.tcl from the RqmRootTreeTools package. Before running the QA code, you should customize this file which is used to setup your job's parameters.
By default, the program will pick up the tcl in your workdir and will ignore the file present in the RqmRootTreeTools directory. You should avoid touching that file to make sure you won't commit your personnal setup in the CVS archive.
Make sure to modify the following parameters before running any job.
- Replace <MAIL:NO> (default option) by <MAIL:YES> to ensure that an e-mail summarizing the run-by-run automated checks is sent at the end of each job. Otherwise, you'll have to scan the logfiles to find the runs which did not pass your full set of automated checks which is not very convenient.
- Later in the file, set all systems but the DIRC to 'NON ACTIVE': <ACTIVE:XXX:NO>, XXX being the subsystem ID. Make sure that the DIRC 'ACTIVE' flag is setup: <ACTIVE:DRC:YES> -- otherwise DIRC QA won't be checked! Keeping only the DIRC active will allow you to save CPU time as by default all subsystems are 'ACTIVE'. Turning off all subystems but the DIRC is harmless as the DIRC tests are independent from the other subsystems.
- Close to the end of the file, you will see the list of experts which will get the automatic e-mails produced by the QA job. If you're not in the list, add your e-mail: <MAILTO:DRC:your e-mail>. To avoid spamming other experts, make sure you commented out all other e-mail addresses!
- Finally, copy the perl macro checkRuns.pl in your workdir. This macro is home-made and aims at launching QA jobs more easily.
If the script completes successfully (it can take quite a long depending on the number of runs to check: check the logfile!), it will produce the following files in the workdir.
Checking the QA of a set of runs requires:
- A logfile called <tag>.rqmlog.
- Two ROOT files ER-new_<tag>.root and ER_<tag>.root which are analyzed in a second step by the ROOT macro plotDircStripcharts.cc.
- A postscript file showing the stripchart evolution 'a la RqmRootTreeTools': ER_<tag>.ps.
- A similar file produced by plotDircStripcharts.cc: DIRC_stripcharts_<tag>.ps. For each stripchart, two plots are displayed: on the left, the histogram of the variable; on the right, its evolution as a function the run number. Red lines show the boundaries of the allowed region for this stripchart.
- Various other files not interesting for QA.
- to run RqmMakeAll of the list of runs as described above;
- to check the QA plots of the runs which don't pass the DIRC automatic monitoring;
[Bad runs are normally clearly visible: they fail most of the automated checks and the corresponding stripcharts often have crazy values (null occupancy etc.)]
- to look at the DIRC QA stripcharts to have an overview of the run quality and to identify unexpected long-term evolution of some stripcharts -- like for instance a net decrease of the mean number of photons per track.
Additional (and clearer?!) information on the way to run RqmMakeAll can be found in the README file of the RqmRootTreeTools package.
This page is maintained by Nicolas Arnaud
Last significant update: November 13th 2005