In this service template (service.yml), the following is created:
- OpenStack Security Group and Keypair
- OpenStack VM
- NiFi instance
- Data pipeline that moves the initial dataset (dataset.tar.xz) from S3 bucket to GridFTP server located in the HLRS HPC testbed
- Execution of HPC job that extracts the dataset, performs test ML training on the dataset and produces the model (protobufs/PeakLens_original.pb)
- Data pipeline that moves the resulting model from HLRS GridFTP server into another S3 bucket
In order to run the test service, the following packages are needed:
- openstack==0.52
- xOpera
Additionally, the following sister directories must exist:
$ tree -L 1
.
├── demo-snow-hpc-training <- this repository
├── iac-modules
└── S3toGridFTPpipeline
- iac-modules:
git clone https://github.com/RADON-SODALITE/iac-modules
- S3toGridFTPpipeline: an extracted and merged CSAR of these service templates: S3toGridFTPpipeline gridFTPtoS3pipeline. Once extracted, the files in the
\_definitions
directory should be modified pointing to relative path for node and relationship types, e.g./nodetypes/radon.nodes.nifi/Nifi/files/stop/stop.yml -> ../nodetypes/radon.nodes.nifi/Nifi/files/stop/stop.yml
/relationshiptypes/radon.relationships.datapipeline/ConnectNifiLocal/files/connect.yml -> ../relationshiptypes/radon.relationships.datapipeline/ConnectNifiLocal/files/connect.yml
To prepare needed node and relationship types, run:
$ make prepare
To deploy, firstly modify inputs.yml and then run:
$ make deploy