Nova Simulation Problem #6018
CMap Error from ND rock files
I'm getting the following error from the official ND rock single files. The are the secondary files for the ND generation. The error is:
nova: /build/nova/novasoft/releases/S14-04-22/CMap/CMap_service.cc:54: void cmap::CMap::postBeginRun(const art::Run&): Assertion `!rdcol.empty()' failed.
This error happens ~30% of the time for Grid jobs.
#2 Updated by Christopher Backhouse over 5 years ago
The particular line you're hitting looks to be dead code. Whatever this xml override mechanism is supposed to be, I don't believe it's ever actually done anything.
So we could rip all that out, but I expect someone else (like Geometry) would throw the same error.
How does one make a Run without a RunData object in it? Especially nondeterministically?
#5 Updated by Christopher Backhouse over 5 years ago
For the hordes of people about to be looking at this, here's why this is mysterious:
The RunData object is put into the event unconditionally in GENIEGen::beginRun().
And then in CMap::postBeginRun() we try to get it out.
There shouldn't be any chance for a race condition. beginRun() comes before postBeginRun(). That's the point.
I recommend printouts to find if/when these functions are called.
#6 Updated by Daniel Pershey about 5 years ago
I ran into this problem when trying to piece together a reco fhicl. While running the reco on a GENIE file in the ND, the job would always crash with the CMap error. After comparing the fhicl to an official reco job, I found I was missing the lines:
physics.analyzers.metadata.params.Reconstructed.base_release: "" #automatically picked up at run time
Putting in these lines fixed the problem somehow.
#7 Updated by Gavin Davies over 4 years ago
This is STILL an issue, 11 MONTHS on...
Ruth and I are running into this for the latest round of simulation processing.
Here's a couple of log files with job specifics and the error:
We really need to nail this one long-term...Out of 10 test jobs, 2 hit this error. Sure not a huge statistical sample but that's still 20% here...