Creation of snapshot for reconstruction (i.e. to reconstruct real data with no grid access)


How can a CCDB snapshot be prepared to allow for a reconstruction w/o internet (grid to be more precise) access ?


@shahoian @zampolli any advice ?

You can run your task once with


which will cache fetched objects locally.
Then, you run with the same settings, the objects will be taken from <locCCDB> directory, but note that this does not support objects changing during the run.
In general, you will need to create a local ccdb server and sync. it with the production server in advance.

Hi @shahoian ,

Your suggestion assumes that at the very beginning there is connection, though. I think that pure without grid is not possible.


how one can create a snapshot w/o connection to ccdb at all?

For the HEP-score benchmark we have followed this procedure:
a) some (CI) machine can produce this snapshot
b) optionally publish it to CVMFS
c) run the workflow against this (CVMFS) snapshot. It works well.

Limitations are (as Ruben pointed out):
*) no time variation in objects
*) calls to CCDB only fetching meta-data (such as RCT/RunInformation) are not in the snapshot for now … but on the other hand also don’t need GRID access

I believe that Laurent expected something like the OCDB that we had with Aliroot - then moved to CVMFS.

Thank you all. Will try.
@zampolli actually no :wink: for now I was indeed asking how to produce the snapshot from a grid-connected machine and then re-use it from a non grid-connected machine.

Good, than I misunderstood.