Prerequisites. For a server to run a Docker container and to serve an API, the following are needed:

• Docker engine. Instruction to install Docker engine for Ubuntu.
• An openned inbound port. In the example here, the openned inbound port is 5003. If the server is an AWS EC2 instance, see here for instruction to open inbound ports.

What does the Docker image contain?

• This Docker image is for easily publishing your algorithms written in Python online, allowing them to be callable through REST api.
• The Docker image has the following packages and their dependencies installed:
  • python=2.7
  • numpy
  • cartopy
  • xarray
  • matplotlib
  • netCDF4
  • scipy
  • pyferret
  • flask
• The Container image is in the Docker hub: opennexfree/opennex_docker .

Your own code

• The files of your own code are stored in a directory outside of the Docker image, and the directory is mounted when the Docker container is launched.
• So this Docker image can be used for any code as long as the following simple rules are observed:
  • The interface to the docker is a file called run.py, which contains up to ten functions from app1() to app10(), corresponding to /app1 to /app10 in the REST api, and performing different algoritms.
  • The parameters passed into those functions are the same as the the parameters specified in the api call.
  • Those functions also take an addictional parameter outDir specifying the directory to store the result.
  • Those functions return a tuple of 3 elements: a text string of text output, a string containing the path of an image file, a string containing the path of a data file. Any of the elements can be None.
  • The REST api returns a json object with 3 key/value pairs, corresponding to the tuple mentioned above: {'message': messageStr, 'plotUrl': PlotUrlStr, 'dataUrl': dataUrlStr} .

What are in this sample code?

• The sample run.py has three functions: app1(), app2() and app3():
  • app1() is a trivial funcation taking in numbers a and b, and returning a+b.
  • app2() takes an argument nc_file specifying the path of a NetCDF file, and returns a summary of the NetCDF file.
  • app3() takes an argument nc_file specifying the path of a NetCDF file, and returns a summary and a plot of the NetCDF file.

How to use it?

• A simple user code can be check out from github:

git clone https://github.com/benyangtang/opennex_sample1.git

• Create a directory myDirToStoreResult to store result.
• Launch the Docker container:

export USER_DIR=/home/ubuntu/opennex_sample1/user
export STATIC_DIR=myDirToStoreResult
docker run -it -d --name nex -p 5003:5000 --rm -v $USER_DIR:/app_dir/user -v $STATIC_DIR:/app_dir/user/static opennexfree/opennex_docker:v01 

Now your Python functions are available to be called by either from a terminal, or from a browser.

A server running the example code can be tested:

• by a command line:

# for app1:
curl 'http://ec2-13-56-153-11.us-west-1.compute.amazonaws.com:5003/app1?a=1&b=10'

# for app3:
curl 'http://ec2-13-56-153-11.us-west-1.compute.amazonaws.com:5003/app3?nc_file=https://podaac-opendap.jpl.nasa.gov:443/opendap/allData/aviso/L4/dynamic_topo_1deg_1mo/zos_AVISO_L4_199210-201012.nc'

• Or from a browser. app1 app3.