(r47) SWGuideUsersFAQ < CMSPublic

CMSPublic Web>SWGuide>SWGuideTroubleShooting>SWGuideUsersFAQ (revision 47) (raw view)~~EditAttachPDF~~
---+!! Questions and answers

%TOC%

#GoAls
This page collects frequently asked questions. Some questions (and answers) concerning common problems have been included here from the Hypernews forums.
You can also submit question through  the [[https://savannah.cern.ch/projects/cmssw/][CMSSW Savannah page]]. Choose
"Submit" from the "Support" menu bar or go directly to [[https://savannah.cern.ch/support/?func=additem&group=cmssw][the Submit page]].

Please check [[http://cms.cern.ch/iCMS/jsp/page.jsp?mode=cms&action=url&urlkey=CMS_US][the existing documentation]], especially
[[WorkBook][the CMSWorkbook]] before adding your question. 

<!--
---+++!! no dictionary for class pair< .... > is available
hn-cms-swDevelopment, Sasha Nikitenko, 31 May 2007

I am reading CMSSW_1_3_1 events from Spring07 production and use line
<verbatim>
HepMC::GenVertex* v = (*p)->end_vertex();
</verbatim>
in analysis code. I got the error message reading events:
<verbatim>
no dictionary for class pair<int,HepMC::GenVertex*> is available
</verbatim>
what can be the problem ?

__Answer:__

Warnings having to do with pair <> are fake (root issue if I am not wrong).
They do not cause any harm.  (Tommaso Boccali)
-->
---+++ InputTag src = "secondCluster"
hn-cms-crabFeedback, Khristian Kotov - Sep 09

I am running CMSSW_1_6_0 and default version on crab on lxplus cluster.
When I create jobs crab generates crab_0_*/job/CMSSW.cfg file which
has two problems:
<verbatim>
line 3885:   InputTag src = "secondClusters"
line 4814:   InputTag src = "thirdClusters"
</verbatim>
Perhaps, crab just uses some corrupted .cfi file from the release.
I tried to fix CMSSW.cfg and submit my jobs, but I still have same
problem in job's output
Does anyone know how to solve such problem?

__Answer:__

I fixed it in my local area by taking from CVS the two files (!RecoParticleFlow/PFTracking/data/second.cff and
!RecoParticleFlow/PFTracking/data/third.cff) and
removing the quotes. (Andrea Rizzi)



---+++ project CMSSW: project: command not found
Help desk, 30 May 2007

An error message type
<verbatim>
project CMSSW
sh: project: command not found
</verbatim>
may appear if you change your shell at lxplus without changing the default shell. 

__Answer:__

See the next question. The problem may arise also when connecting to lxplus from a Windows
PC and opening a second xterm. The environmental variables do not get properly set in this case, use
the original x-window, instead. It may also happen for a similar reason if you have saved your profile
and you are not prompted for you password when opening a lxplus window on Windows PC.

---+++ How to change the default shell on lxplus
Help desk, Dayong Wang, 30 Apr 2007

How to change the default shell on lxplus?

__Answer:__

You can find the instructions in the CERN IT list of frequently asked question: [[http://consult.cern.ch/qa/1599]].

---+++ Copying a small amount of data locally
cms-User.Support, Michel Della Negra, 29 Mar 2007

In trying to execute
<verbatim>
rfcp /castor/cern.ch/cms/store/h2tb2006/reco/v6/h2.00030109.combined.OutServ_0.0-cmsswreco.root tbh2.root
</verbatim>
I got: 
<verbatim>
Disk quota exceeded.
</verbatim>
The file contains 50K events from TB2006 H2 data.
I do not need the full statistics  to debug an analysis.
Is there an exec file for copying only say 1000 events?

__Answer:__

You can copy a fraction of a data set from the local storage (i.e. from castor at CERN) by running cmsRun with 
the configuration file which just copies a part of the file locally as explained in WorkBookDataSamples#LocalCopy.

You can skip a number of events by adding =untracked uint32 skipEvents = 100= (where 100 would be the number of events to be skipped)
in the !PoolSource block and you can start with a specified event by adding
<verbatim>
   untracked uint32 firstRun = 5006
   untracked uint32 firstEvent = 96
</verbatim>
in the !PoolSource block. (Luca Lista, Chris Jones, Iana Osborne)


---+++ CLHEP::HepLorentzVector
hn-cms-progQuestions, Carmelo Marchica, Mar 12 2007

I would like to use the following variable pt:

<verbatim>
std::vector<HepMC::GenParticle*> electrons; double pt = electrons[0]->momentum().pt()
</verbatim>

And I get: 
<verbatim>
no matching function for call to `CLHEP::HepLorentzVector::pt()&#8217;
</verbatim>

What I have to use for the pt? And where I can find the definitions of the variables in momentum()? 

__Answer:__

You need to use =momentum().perp()= . You can check the interface of 
CLHEP::HepLorentzVector from (for example) LXR: 
[[http://cmslxr.fnal.gov/lxr/source/external/clhep/1.9.2.2/include/CLHEP/Vector/LorentzVector.h]] (%RED% 22 Mar 07: the link temporarily broken due to LXR %ENDCOLOR%) or check the [[http://proj-clhep.web.cern.ch/proj-clhep/][CLHEP documentation]] (not necessarily very helpful). (Andrea Bocci)

---+++  No "EcalWeightXtalGroupsRcd" record found -> !FakeConditions.cff
hn-cms-ecalDB, Ian Tomalin, Mar 03 2007

I am running the b-jet HLT code with CMSSW 1.3.0.pre5. In order to
reconstruct jets in the calorimeters (starting from Digis), the
configuration file includes:

<verbatim>
include "HLTrigger/Configuration/data/common/CaloTowers.cff"
include "HLTrigger/Configuration/data/common/RecoJetMET.cff"
</verbatim>

which in turn include:

<verbatim>
include "RecoLocalCalo/Configuration/data/RecoLocalCalo.cff"
</verbatim>

When I run this job, I get the error message:

<verbatim>
========================================================================
No "EcalWeightXtalGroupsRcd" record found in the EventSetup.
Please add an ESSource or ESProducer that delivers such a record.
cms::Exception going through module
EcalWeightUncalibRecHitProducer/ecalWeightUncalibRecHit.
========================================================================

</verbatim>
 I didn't get this error with pre2. I am not an ECAL expert. How do I get rid of this error ?

__Answer:__

Starting in pre4 you need to add to your your cfg

<verbatim>
include "Configuration/StandardSequences/data/FakeConditions.cff"
</verbatim>

if you are going to to reco. (Shahram Rahatlou)





---+++ !SuperCluster->e3x3(), Invalid Reference
hn-cms-physTools, James Jackson, Mar 03 2007

I am using MC data (FEVT) generated with 1_2_0, analysing with 1_2_1.  
Getting the supercluster associated with a !PixelMatchGsfElectron is  
working fine, and the energy() method returns sensible values.  
However, if I attempt to call e3x3() on the supercluster, I get the  
following exception:

<verbatim>
%MSG-w ScheduleExecutionFailure:  PostModule 03-Mar-2007 18:13:08 GMT
        Run: 3766 Event: 46
an exception occurred and event is being skipped:
---- ScheduleExecutionFailure BEGIN
ProcessingStopped
---- InvalidReference BEGIN
NullID RefCore::getProduct: Attempt to use a null Ref.cms::Exception  
going through module L1TriggerEfficiencyAnalyzer/analysis run: 3766  
event: 46
---- InvalidReference END
Exception going through path p
---- ScheduleExecutionFailure END
</verbatim>

Looks like something fishy going on - is this a known bug /  
incompatability between 1_2_0 and 1_2_1?

__Answer:__

This is a known issue, and the e3x3() and other cluster shape  
accessor methods are removed from the !SuperCluster interface from  
130_pre3 onwards.  These methods do not to work because the Ref to  
the !ClusterShape is not set and it was decided instead to set up an  
association map.  Therefore to access shape information it is  
necessary to use the association map between !BasicClusters and  
!ClusterShape, for example:

<verbatim>
edm::Handle<reco::BasicClusterShapeAssociationCollection>  barrelClShpHandle;
iEvent.getByLabel("hybridSuperClusters","hybridShapeAssoc",  barrelClShpHandle);

edm::Handle<reco::BasicClusterShapeAssociationCollection> endcapClShpHandle;
iEvent.getByLabel("islandBasicClusters","islandEndcapShapeAssoc", endcapClShpHandle);

reco::BasicClusterShapeAssociationCollection::const_iterator  
seedShpItr = barrelClShpHandle->find(myBarrelSuperCluster->seed());
const reco::ClusterShapeRef& seedShapeRef = seedShpItr->val;
float e9 = seedShapeRef->e3x3();
</verbatim>
(David Futyan)


<!--
---+++!! RFIO problem when reading a large number of files in one job
hn-cms-edmFramework, Patrick Janot, Feb 22 2007

I am trying to read the single pion files fully simulated for the HCAL response tuning. Each sample is dedicated for a single pion energy (1,2,3,5,9,15,20,30,50,150,225 and 300 GeV), and contain between 70 and 140 files in castor.  Very often, in a random manner, 
my jobs just stop with an rfio error, as reported below. The incriminated file changes from one run to the other. While this happened 
10% of the times last week, this fraction increased to 100% this week, hence preventing me to work (none of my jobs come to an end, 
and I never have my output root tree...).
<verbatim>
<FrameworkError ExitStatus="8000" Type="SEALException" >
  seal::Exception caught in cmsRun
I/O operation rfio_open(rfio:///?path=/castor/cern.ch/cms/store/mc/2006/12/26/mc-hcal-120_Ieta1-29_Iphi1-2_E020/0001/FE6934BE-42B7-DB11-95A5-0050455D17CC.root,0,438) failed RFIO e
rror 0/22: Invalid argument
</FrameworkError>
</verbatim>

Could you help me, and tell me what I should do to avoid these crashes (independent of my code)?

__Answer:__

No solution, yet, but I ad-hoc recipe by Patrick can be found in
[[https://hypernews.cern.ch/HyperNews/CMS/get/edmFramework/622.html]]
-->
---+++  !MagGeometry::fieldInTesla: failed to find volume for (nan,nan,nan) 
hn-cms-swdevelopment, Javier Fernandez, 12/11/2006

I&#180;m experiencing randomly in some jobs over the grid the following
error when doing re-reconstruction with CMSSW_1_0_6 over CSA06 1_0_6
samples (I&#180;m not saying this is related to this samples/version):
<verbatim> 
MagGeometry::fieldInTesla: failed to find volume for (nan,nan,nan)
</verbatim>
 the stdout log file is filled with this line at some random event 
until the maximum output log file size is reached, while the stderr log 
 file is filled with the following line:
<verbatim> 
  *** Break *** write on a pipe with no one to read it
</verbatim>
resulting in 2x2Gb std log files. Other jobs run good on the same sample

__Answer:__

The warning means that field has been requested for the position 
(nan,nan,nan) which is obviously invalid (nothing to do with the field 
itself!).

Any piece of !CMSSW can cause this printout - a rare occurrence of a 
numerical proble, creates a nan that is propagted in the code and 
triggers this message when it reaches the field.

(it should be quite worring that apart of this warning message, jobs go 
on and the nan is silently accepted...)

This particular issue with 106 is solved as mentioned by Chang, but 
there can be more such problems in other parts of !CMSSW.

the way to trace down the source of the problem is to add:

<verbatim> 
service = EnableFloatingPointExceptions {}
service = Tracer {}
</verbatim>
that will cause an exception to be thrown at the place where the nan is 
created, allowing to debug it. This does not help Javier: the 
application will normally crash, unless there is a way to instruct the 
framework to skip events with such an exception.

I think that it would be a good idea to require release validation to be 
run with !EnableFloatingPointExceptions so that these bugs are found 
early in the release cycle.

What I can do on the magnetic field side is, on the other hand, one (or 
more) of the following:

   1 print the message up to a given number of times
   1 throw an exception (will force people to debug their code), possibly printing the stack trace so the culprit can be found easily
   1 stop complaining and just go on (as all the rest of !CMSSW does ;-)

This will happen in 130 I hope. (Nicola Amapane)

---+++ Naming of the samples in DBS
hn-cms-btag, Pavel Demin, 07 February 2007 

Could someone please explain me what QCD samples are intended to be used for b-tag validation? I can see that currently there are 
two types of samples: 
   * mc-onsel-120_PU_QCD_pt_80_120
   * mc-onsel-120_QCD_pt_80_120
I could not find any description of the samples in DBS.

__Answer:__

  In http://cmsdbs.cern.ch/discovery/ :

   1 Datasets with "onsel" in the name were produced at the request of the HLT group. The corresponding configuration files are in the CMSSW directory =Configuration/CSAProduction/data/=.
   1 Datasets with "physval" in the name were requested by the Physics Validation Group. The .cfg files are in =Configuration/PhysicsValidation/data/=.
   1 Datasets with "relval" in the name were requested by the Software Validation Group. The .cfg files are in =Configuration/ReleaseValidation/data/=.

The =data/= directories all contain a subdirectory =PileUp/=, which contains any .cfg files used for pileup production.

  (In some cases, you may need to check out the CVS HEAD of
Configuration to find them ...)

  So in your example: =mc-onsel-120_PU_QCD_pt_80_120= was produced with
=Configuration/CSA06Production/data/PileUp/CSA06Production/data/PileUp/QCD_pt_80_120_PU_OnSel.cfg= , whilst =mc-onsel-120_QCD_pt_80_120= was produced
with =Configuration/CSA06Production/data/QCD_pt_80_120_OnSel.cfg= . And
our pileup bb and cc dijets are produced with
=PhysicsValidation/data/PileUp/BBbar_pt80to120_LowLumi.cfg= etc. (Ian Tomalin)



---+++ Compatibility between !CMSSW versions
hn-cms-edmFramework,  Pedrame Bargassa, 23 Nov 2006

Is there any known 120pre2 -> 106 incompatibility as well ?

__Answer:__

Yes, you should expect to read the CSA06 MC data only up to CMSSW_1_0_6.

  The general picture is:

| *Type of data* | *produced in* | *run in* | 
| GEN-SIM-DIGI  | 0_8_x-1_0_x |1_0_x (preferably 1_0_6) |
| GEN-SIM-DIGI-RECO |1_0_x   |1_0_x (preferably 1_0_6) |
| GEN-SIM-DIGI      | 1_1_1       |1_1_x |
| GEN-SIM-DIGI-RECO | 1_1_1  |1_1_x |
 
i.e. in general don't expect to read any MC data written in 1_0_x with 
1_1_x and don't expect to read anything written in 1_1_x with 1_2_x. (It 
may run, but produce garbage results, don't waste your time on it.)

 Right now it looks like only GEN-SIM data will be usefully compatible
between 1_1_x and 1_2_x. (Peter Elmer)

NB, this table has been updated in WorkBookWhichRelease

---+++ How to access trajectories and track state on each crossed layers
hn-cms-swDevelopment, Boris Mangano, 18 Oct 2006

In the past, there were many requests to access track's parameters (in
particular the complete !TrajectoryMeasurement) on the surface of each
layer crossed by the track. The trajectory is the object devoted to
collect all the !TrajectoryMeasurements (TM) of the tracks.
Nevertheless !TrajectoryMeasurement objects were not designed to be saved
on disk and, after the final fit of the tracks is done, the
informations about the TMs is discarded. Final Track contain only the
TSOS on the innermost and outermost hits.

__Answer:__

Since last nightly and, hopefully, in120pre2:
Thanks to a very recent update of the Framework, there is the
possibility to define transient-only product which can be put into the
event, even if they are not allowed to be saved on the
root file.

The !RecoTracker/TrackProducer has been accordingly modified to be able
to put the vector of trajectories into the event. The default behavior,
for saving cpu time, is to discard the copy. Nevertheless,the user can
now change the new !TrackProducer's parameter "TrajectoryInEvent" from
'false' to 'true' and the trajectories will be put into the Event during
the execution of the job.

A subsequent module can get them back from the event using the standard
commands:
<verbatim>
Handle<vector<Trajectory> > trajCollectionHandle;
iEvent.getByLabel(param_.getParameter<string>("trajectoryInput"),trajCollectionHandle);
</verbatim>
and then analyze them:
<verbatim>
for(vector<Trajectory>::const_iterator it =
trajCollectionHandle->begin(); it!=trajCollectionHandle->end();it++){
      edm::LogVerbatim("TrajectoryAnalyzer") << "this traj has " <<
         it->foundHits() << " valid hits" ;

vector<TrajectoryMeasurement> tmColl = it->measurements();
.............
</verbatim>
A complete analyzer (=TrajectoryAnalyzer .cc .cfg=) has been committed in
the [[http://cmssw.cvs.cern.ch/cgi-bin/cmssw.cgi/CMSSW/RecoTracker/TrackProducer/test/?cvsroot=CMSSW][RecoTracker/TrackProducer/test]] directory. It explains how the value
of the TMs can be accessed.

Last remarks, the trajectories can be accessed after the !TrackProducer
module is executed (with its default !TrajectoryInEvent parameter
changed) or after the Refitter module: this second producer, has also
the new !TrjectoryInEvent parameter.

The important thing is that the 2 modules (!TrackProducer+"user defined
trajectory analyzer" or !TrackRefitter+"user defined..") are run in the
same job. The trajectories are not persistent! (Boris Mangano)

---+++Can't compile an EDAnalyzer created with =mkedanlzr=
hn-cms-swDevelopment, Benedict Huckvale, 10 Oct 2006

I've done nothing more than
<verbatim>
> eval ...
> mkedanlzr WGAnalyser
> rm -r interface/ test/ doc/
</verbatim>
But when I try to build I get this:
<verbatim>
benedict@lxplus061:~/w0/cmssw/CMSSW_1_0_1/src/WGAnalyser > scramv1 b -r
Resetting caches
Parsing BuildFiles
..
 /afs/cern.ch/user/b/benedict/w0/cmssw/CMSSW_1_0_1/tmp/slc3_ia32_gcc323/Makefile:536: *** target file
`src_clean' has both : and :: entries.  Stop.
</verbatim>

__Answer:__

You need to put the generated package into a subsystem (i.e. NOT under
<project>/src/myAnalyzer but <project>/src/subsys/myAnalyzer. (Shaun Ashby)
<br/>
Unfortunately you might also find you need to check out an entirely new CMSSW project area.
Try not to build a wrongly placed package by accident in future! (Benedict Huckvale)

#NoReco
---+++Problems when removing =RECO.cff= from a config file
hn-cms-swDevelopment Taylan Yetkin, Sep 23, 

I would like to remove reconstruction part from the others. When I remove
=include "Configuration/Examples/data/RECO.cff"=
and reconstruction from path
=path p = {smear_and_geant,digitization, reconstruction}=
which are in the exampleRunAll.cfg my naive guess is that it should  
work. However it does not.

__Answer:__

I have found what in =RECO.cff= depends on the !SWGuideParticleDataTable: it  
is the =HepMCCandidateProducer= which is then used by different Jet and  
Met algorithms to do their work on the mc particles instead of the  
reconstructed quantities.  So I guess =RECO.cff= is not a completely  
accurate name, it probably should be called =MCRECO.cff= since it is  
designed to work with MC and not actual data.
To solve your problem, I suggest adding

 =include "SimGeneral/HepPDTESSource/data/pdt.cfi"=

to your config file.
(Chris Jones)

#MoreThan1File
---+++ Problems accessing more than one file 

hn-cms-swDevelopment,  Christian Veelken, Wed, 13 Sep 2006


I have two Monte Carlo files that were produced independently and each
contain 500 events. I have written a class derived from EDAnalyzer to make
plots of the file content. When I run over one of the two files at a time,
everything is fine, no problem. But when I run over both MC files in the
same cmsRun job, cmsRun crashes when processing the first event in the
second file.

__Answer:__

_Update 16 Apr 07: A description and an example how to save histograms or other objects from different modules in one ROOT file is available in SWGuideTFileService_ 

My guess is the program crashes in
your EDAnalyzer.  Are you by chance filling your own ROOT file from
within the EDAnalyzer?  If so, then other have found that when the
!PoolSource goes from one input ROOT file to another one their code
crashes.  What was determined is unless you've explicitly openned
your own ROOT file and tied your histogram, ntuples, etc. to your
file, then ROOT will tie those things to the first 'input' file and
when that input file goes away the pointers to your histograms,
ntuples, etc. then point to object that no longer exist (since ROOT
deleted them when the file closed).   So it look like you need to
modify the ROOT calls you are making in your EDAnalyzer.
(Chris Jones)

If Chris' diagnosis is correct (very likely), you can have a
look at
[[http://cmssw.cvs.cern.ch/cgi-bin/cmssw.cgi/CMSSW/RecoEcal/EgammaClusterProducers/src/EgammaSimpleAnalyzer.cc?rev=1.6&cvsroot=CMSSW][RecoEcal/EgammaClusterProducers/src/EgammaSimpleAnalyzer.cc]]
where I had a similar problem and is now solved by correctly
binding the histos and trees to my root file w/o interfering
with !PoolSource.
(Shahram Rahatlou)


#FindData
---+++ Where can I find the data?

How do I know what data is available (real data from MTCC or simulation data) and how can I access it?

__Answer:__

See [[WorkBookDataSamples][the Workbook page on data access]].

For the MTCC data (cosmic muons at CMS) see the instructions in [[IGUANA][the IGUANA web page]],
and in particular [[EventDisplayShifterIntroduction#MTCCDataFiles][the data file location]].

The CSA06 samples can be found using [[http://cmslcgco01.cern.ch:8001/][DBS]] tool.

For each release of the !CMSSW packages, a set
of of small event samples are produced which are mainly used to validate the release.
Find the instructions how to access this data in [[RelValSamples][the release validation samples page]].
 
<!--#AddCrossSection
---+++!! How to save cross section in the root file?
Sami Lehti, 21 Jul 2006

In an analysis containing several channels (signal+backgr), it is very helpful 
if some normalization constants, like cross section and preselection efficiency, 
can be stored in the root file.
How is this done in practise (a working example would be nice)? How is the 
cross section read from the root file? Can the stored cross section be 
changed by hand afterwards? For example if i have 1M events in root file(s), 
and i get a new cross section from theorists, do i have to rerun the 1M events,
or can i change the cross section value in the root file(s)? If yes, how?

__Answer:__

The mechanism to add a cross-section to the run or job (i.e. one number per run) is not yet ready, it is planned but not yet there (it is foreseen by the end of August, at the earliest)

Meanwhile, you can try the following solution which [[UserSupportNumberToEachEvent][writes a number to each event]].-->




-- Main.KatiLassilaPerini - 13 Jul 2006