SHERPA is hosted by Hepforge, IPPP Durham

Running Sherpa docker containers

As of version 2.2.2 we provide docker images on dockerhub (https://hub.docker.com/u/sherpamc/)

A basic version comes with rivet and ufo support while a more advanced build comes with Openloops and LHAPDF.

Both images come with mpirun support.


Obtaining docker images

The images can be pulled from dockerhub as such

docker pull sherpamc/sherpa-basic:2.2.2

for the basic version and

docker pull sherpamc/sherpa:2.2.2

for the version that comes with Openloops and a selection of loop processes.

Running docker images

After pulling an image Sherpa can be used immediately through an interactive shell. A simple example would be

docker run -it sherpamc/sherpa-basic:2.2.2 bash

This command 'boots' the docker image and uses bash as entrypoint.

As Sherpa is installed in /usr/local Sherpa could for instance be invoked from within the docker container as such:

mpirun -n 6 Sherpa /usr/local/share/SHERPA-MC/Examples/Jets_at_LeptonColliders/LEP_Jets/Run.dat

Running with local folders mounted

To make e.g. the current directory on the host system available in the Docker container, the following command can be used:

docker run -it -v $PWD:$PWD -w $PWD sherpamc/sherpa-basic:2.2.2 bash

If the current directory contains a run card, the following command will invoke Sherpa directly and all output files will be readily available on the host system:

docker run -it --rm -v $PWD:$PWD -w $PWD sherpamc/sherpa-basic:2.2.2 Sherpa -f Run.dat EVENTS=1000

In cases where you get error messages such as

Sherpa: Process_Group::ConstructProcesses throws fatal error: 
   Cannot open '/mt/home/hschulz/Docker/MCnetDocker/Sherpa/release/SHERPA-MC-2.2.2/bl2/workdir/Process/Sherpa/2_2__j__j__l__l.map'

the docker user won't have write permissions in the current directory. This can be overcome by specifying the user id as such:

docker run -it -u `id -u $USER` --rm -v $PWD:$PWD -w $PWD sherpamc/sherpa-basic:2.2.2 Sherpa -f Run.dat EVENTS=1000

Generate UFO model and run custom rivet analysis

This example shows how to firstly run Sherpa-generate-model through docker and how to build and use a custom rivet analysis (i.e. one that is not included in Rivet2.5.3)

docker run -it  --rm -u `id -u $USER`  -v $PWD:$PWD -w $PWD sherpamc/sherpa-basic:2.2.2 Sherpa-generate-model --installdir $PWD MY_UFO_MODEL

The model library will be installed in the hosts $PWD.

docker run -it  --rm -u `id -u $USER`  -v $PWD:$PWD -w $PWD sherpamc/sherpa-basic:2.2.2 rivet-buildplugin RivetNewAnalysis.so NewAnalysis.cc

Both commands assume that the model folder as well as the rivet analysis code are present in the hosts $PWD.

To set the RIVET_ANALYSIS_PATH and run Sherpa with the new model --- assuming that the corresponding (analysis) block was added to the Example runcard, i.e. that this is present in the runcard

(analysis) {
BEGIN_RIVET {
-a NewAnalysis
} END_RIVET
}(analysis)

--- Sherpa can be invoked as such:

docker run -it -e "RIVET_ANALYSIS_PATH=$PWD" --rm -u `id -u $USER`  -v $PWD:$PWD -w $PWD sherpamc/sherpa-basic:2.2.2 Sherpa -f MY_UFO_MODEL_Example_Run.dat ANALYSIS=Rivet 

Simplification of command

For convenience, an alias can be created as such

alias Sherpa='docker run -it -e "RIVET_ANALYSIS_PATH=$PWD" --rm -u `id -u $USER`  -v $PWD:$PWD -w $PWD sherpamc/sherpa-basic:2.2.2 Sherpa'

Doing so allows to run the above example in this way:

Sherpa -f MY_UFO_MODEL_Example_Run.dat ANALYSIS=Rivet
Last modified 4 weeks ago Last modified on Jan 26, 2017, 7:29:36 PM