This is basic setup for running the ParFlow container image. Running both
Please report issues at the ParFlow issue tracker: https://github.com/parflow/parflow/issues
The container image should work with any of the popular systems. Docker is the most popular but Podman and Singularity have also been used to run this image.
Documentation for downloading and installing these container systems is available at: https://docs.docker.com/ , https://podman.io/ and https://sylabs.io/
This example repository has scripts and input for running a simple ParFlow example under Docker.
If you have git installed, you can clone the example:
git clone https://github.com/parflow/docker.gitIf you do now have git, you can download a zip file of the example:
https://github.com/parflow/docker/archive/master.zip
Running ParFlow using the container instance is similiar to how ParFlow is run from the command line using Python or TCL. The python/tclsh command is replaced with a platform specific script.
Where you see:
tclsh <arguments>python <arguments>to run with the the container instance use the provided script:
./bin/parflow.sh <arguments>On Windows the script to use is parflow.bat.
The first time you execute the 'parflow.bat' or 'parflow.sh' script Docker will download the ParFlow image. This requires an internet connection and may take several minutes.
The file extension of the ParFlow input (.py or .tcl) will be used to determine which script engine to use.
To execute the default_single Python test case provided with this download:
./bin/parflow.sh default_single.py 1 1 1To execute the TCL default_single TCL test case provided with this download:
./bin/parflow.sh default_single.tcl 1 1 1This wrapper script is invoked as one would invoke TCL outside of docker. In this case we supply the TCL ParFlow input file (which is always required) and the core topology (in this case a 1 core along the X/Y/Z axis).
To run this same problem in parallel:
./bin/parflow.sh default_single.tcl 2 1 1Will split the problem in the X axis across two cores.
To execute the default_single test case run with
.\bin\parflow.bat default_single.tcl 1 1 1This wrapper script is invoked as one would invoke TCL outside of docker. In this case we supply the TCL ParFlow input file (which is always required) and the core topology (in this case a 1 core along the X/Y/Z axis).
To run this same problem in parallel:
.\bin\parflow.bat default_single.tcl 2 1 1Will split the problem in the X axis across two cores.
Docker needs to be configured to run on multiple cores under Windows. In the Docker Desktop go to the Settings -> Advanced tab to set the number of CPU's docker will use. This will be a small number, normally 1 or 2, by default.
This panel also controls the memory allocated for the Docker instances. The default memory is typically 1 or 2 GB. For large problems this will need to be increased.
Windows Docker support currently has issues with file ownership. The files created when will be owned by the Administrator account. The files are readable by you and you can delete files BUT directories can only be deleted by an Administrator account.
We are looking at how to better manage this issue; this is a known 'feature' of Docker on Windows.
The output of running the example will be placed in the example directory.
Below is a list of files from a run done on windows. The '.txt' and '.log' files contain summary information about the turn. The '.silo' and '.pfb' files are the same output, simply stored in two different formats. The Silo files are viewable using Visit (https://wci.llnl.gov/simulation/computer-codes/visit/). See the ParFlow user manual for more information on output.
Mode LastWriteTime Length Name
---- ------------- ------ ----
d----- 7/17/2019 9:19 AM .git
d----- 7/17/2019 9:02 AM bin
d----- 7/16/2019 3:30 PM correct_output
d----- 7/17/2019 9:09 AM default_single.out
-a---- 7/17/2019 9:11 AM 40 .gitignore
-a---- 7/17/2019 9:19 AM 184 default_single.out.concen.0.00.00000.pfsb
-a---- 7/17/2019 9:19 AM 2765 default_single.out.concen.0.00.00000.silo
-a---- 7/17/2019 9:19 AM 604 default_single.out.concen.0.00.00001.pfsb
-a---- 7/17/2019 9:19 AM 2765 default_single.out.concen.0.00.00001.silo
-a---- 7/17/2019 9:19 AM 1324 default_single.out.concen.0.00.00002.pfsb
-a---- 7/17/2019 9:19 AM 2765 default_single.out.concen.0.00.00002.silo
-a---- 7/17/2019 9:19 AM 1804 default_single.out.concen.0.00.00003.pfsb
-a---- 7/17/2019 9:19 AM 2765 default_single.out.concen.0.00.00003.silo
-a---- 7/17/2019 9:19 AM 2364 default_single.out.concen.0.00.00004.pfsb
-a---- 7/17/2019 9:19 AM 2765 default_single.out.concen.0.00.00004.silo
-a---- 7/17/2019 9:19 AM 2724 default_single.out.concen.0.00.00005.pfsb
-a---- 7/17/2019 9:19 AM 2765 default_single.out.concen.0.00.00005.silo
-a---- 7/17/2019 9:19 AM 2718 default_single.out.concen.press.silo
-a---- 7/17/2019 9:19 AM 5208 default_single.out.log
-a---- 7/17/2019 9:19 AM 17380 default_single.out.perm_x.pfb
-a---- 7/17/2019 9:19 AM 2742 default_single.out.perm_x.silo
-a---- 7/17/2019 9:19 AM 17380 default_single.out.perm_y.pfb
-a---- 7/17/2019 9:19 AM 2742 default_single.out.perm_y.silo
-a---- 7/17/2019 9:19 AM 17380 default_single.out.perm_z.pfb
-a---- 7/17/2019 9:19 AM 2742 default_single.out.perm_z.silo
-a---- 7/17/2019 9:19 AM 7991 default_single.out.pftcl
-a---- 7/17/2019 9:19 AM 17380 default_single.out.porosity.pfb
-a---- 7/17/2019 9:19 AM 2709 default_single.out.porosity.silo
-a---- 7/17/2019 9:19 AM 17380 default_single.out.press.00000.pfb
-a---- 7/17/2019 9:19 AM 713 default_single.out.timing.csv
-a---- 7/17/2019 9:19 AM 4327 default_single.out.txt
-a---- 7/17/2019 9:19 AM 1453 default_single.out.wells
-a---- 7/17/2019 9:19 AM 5598 default_single.pfidb
-a---- 7/17/2019 8:51 AM 13518 default_single.tcl
-a---- 7/17/2019 9:07 AM 1071 Dockerfile
-a---- 7/17/2019 8:51 AM 2972 pftest.tcl
-a---- 7/17/2019 9:19 AM 3229 README.md
Running your input with the Docker has one restriction using the provided script; all input files must be contained in the current working directory. This includes the initial TCL input file and all files that are used in the problem. Subdirectories in the current working directory are fine and will be available in the instance. Soft links will not work.
Once you have all the files in the current working directory simply invoke the script with the TCL file. For example on Windows:
.\bin\parflow.bat my_problem_setup.tclIt is possible to share other directories with the cointainer instance but that is beyond the scope of this document. You will need to bind additional directories by changing the Docker command line. See Docker documenation for information on sharing host file systems with the Docker instance.
By default the supplied script will download and execute the latest version ParFlow. Other released versions of ParFlow are available on DockerHub (starting with version 3.5.0). To run a specific version of ParFlow edit the script to change the version to match your desired version, for example to run version 3.12.0:
docker run -v %cd%:/data parflow/parflow:version-3.12.0 %1 %2 %3 %4 %5 %6 %7 %8 %9