PromptReco failure PromptReco_Run381379_ParkingSingleMuon4 #45162
Comments
|
cms-bot internal usage |
|
A new Issue was created by @Dr15Jones. @antoniovilela, @sextonkennedy, @smuzaffar, @makortel, @rappoccio, @Dr15Jones can you please review it and eventually sign/assign? Thanks. cms-bot commands are listed here |
|
The job can be run by setting up a CMSSW_14_0_7 area, downloading the tarball (which is at Then after untarring go to directory job/WMTaskSpace/cmsRun1 and then do |
|
There appear to be lots of extraneous exceptions being thrown (and caught) in this job. The first one encountered is Which is caught here cmssw/RecoTracker/FinalTrackSelectors/plugins/SingleLongTrackProducer.cc Lines 158 to 173 in dbbd44f which is problematic as the tracks are the |
|
assign tracking |
|
The next group of exceptions come from the exception originates here cmssw/TrackingTools/TrajectoryState/src/PerigeeConversions.cc Lines 15 to 16 in dbbd44f and is caught here |
|
assign reconstruction |
|
New categories assigned: reconstruction @jfernan2,@mandrenguyen you have been requested to review this Pull request/Issue and eventually sign? Thanks |
cmsbuild
added
reconstruction-pending
pending-signatures
and removed
pending-assignment
labels
|
By skipping the first events, I was able to get to the trackback for the exception which ultimately ended the job |
|
assign root |
|
@pcanal how can we understand better what happened during the read? |
|
type root |
|
type tracking |
that's just looks like a poorly written code, where try/catch is used instead of checking for trackExtra to be present. cmssw/RecoTracker/FinalTrackSelectors/plugins/SingleLongTrackProducer.cc Lines 133 to 136 in dbbd44f a proper copy is made conditionally, while the rest in |
|
@borzari Please clarify if you are available to check this. |
|
Hi @slava77 I applied what you suggested in this commit, used the opportunity to remove some duplicated code, and tested it with RelValZMM and RelValTTbar events by comparing the version with try/catch results with the version with the validity check results. Everything worked as intended and no changes to the output were observed, as expected. Just to clarify two points:
|
I misread the TrackExtraBase; So, I would add this Even though in the current setup a track without an extra is enough, there can still be cases where |
Out of curiosity why is that? Can't the |
Hi @mmusich
The hit checks are to make sure that this track won't have missing layers with measurement, which is not 100% effective as I already showed during the presentations about this topic, but also doesn't impact a lot on the final result because it doesn't happen so often. I wouldn't think changing that part of the code for Here I added the suggestions from @slava77. Again, I tested with RelValZMM and RelValTTbar events, and things are working as expected. If you don't have other suggestions, I can open a PR with it and we can continue the discussion there |
Exactly, can't you do that before filling the vector? Default constructed tracks can't be used for refit. |
Alright, so instead of only getting the track with the smallest |
Right, this is what I had in mind. |
It didn't work. If I move the validity check from the rechits/hitpattern check to where I select tracks (I did |
Isn't |
Should be. I implemented it like Slava suggested here Could it be that, although I am adding only tracks with valid |
|
The check I used was |
Alright @Dr15Jones, but does it happens every time I am using a Well, in any case, I would suggest to open a PR with these changes. At least to remove the |
maybe I am missing something, but with CMSTrackingPOG@5318549 on top of borzari@95ecc4b I can run this test: (even using the whole input file) without crashes. |
|
@mmusich most probably I was missing something. The main differences I see (besides the better organization of the code in the way you wrote), is that I included |
|
@mmusich I started from your branch and tested what I mentioned above:
May I start a PR to include your changes and the |
here it is: #45213. I used the CMSTrackingPOG VO so you should be able to push more commits if necessary. |
Great! I don't think there are any other modifications that are needed. Just FYI, I also checked the output DQM histograms of that branch using RelValZMM events and they are the same as before the changes, as expected |
|
@Dr15Jones Is my understanding correct? |
|
@slava77 I'm on vacation until Thursday. The try/catch fixes were only there to make it easier to get to the underlying problem in the debugger. It does look like the underlying problem is in ROOT. |
|
Coming back to problem itself, in https://cms-talk.web.cern.ch/t/paused-job-for-promptreco-run381379-parkingsinglemuon4/42082/7 the likely cause was mentioned to be a corrupted file. I suppose there were no further similar failures? Under the file corruption hypothesis, maybe we could just close the issue? |
|
+1 |
cmsbuild
added
reconstruction-approved
fully-signed
and removed
reconstruction-pending
pending-signatures
labels
|
This issue is fully signed and ready to be closed. |
|
@cmsbuild, please close |
github-project-automation
bot
moved this from Work in CMS
to Done
in ROOT prioritization
From https://cms-talk.web.cern.ch/t/paused-job-for-promptreco-run381379-parkingsinglemuon4/42082
The tarball can be found here:
/afs/cern.ch/user/c/cmst0/public/PausedJobs/Run2024E/FileReadError/job/WMTaskSpace/cmsRun1
From the logs it seems to crash at event 1742503164. The error is reproducible locally.
The text was updated successfully, but these errors were encountered: