HyperFlow: a scientific workflow management system

Description

HyperFlow is a Workflow Management System (WMS) dedicated for scientific workflows.

Browse the wiki pages to learn more about the HyperFlow workflow model.

Getting started

Installation

• Install Node.js (http://nodejs.org)
• Install Redis (http://redis.io)
• Install HyperFlow:
  • From npm package: npm install -g @hyperflow/hyperflow
  • From github repository:<br>npm install https://github.com/hyperflow-wms/hyperflow/archive/{version}.tar.gz<br>(where {version} is for example v1.5.0)
  • From the master branch: <br>npm install https://github.com/hyperflow-wms/hyperflow/archive/master.tar.gz

Running locally

• Start the redis server: redis-server
• Run example workflows using command hflow run <wf_directory>, for example:<br>hflow run ./examples/Sqrsum

Running locally using Docker images

• Use the latest Docker image for the HyperFlow engine, published in Docker Hub as hyperflowwms/hyperflow
• You can build the image yourself: make container
• Start redis container:<br> docker run -d --name redis redis --bind 127.0.0.1
• Run workflow via HyperFlow container, for example:

docker run -a stdout -a stderr --rm --network container:redis \
       -e HF_VAR_WORKER_CONTAINER="hyperflowwms/soykb-workflow-worker" \ 
       -e HF_VAR_WORK_DIR="$PWD/input" \ 
       -e HF_VAR_HFLOW_IN_CONTAINER="true" \
       -e HF_VAR_function="redisCommand" \
       -e REDIS_URL="redis://127.0.0.1:6379" \
       --name hyperflow \
       -v /var/run/docker.sock:/var/run/docker.sock \
       -v $PWD:/wfdir \
       --entrypoint "/bin/sh" hyperflowwms/hyperflow -c "apk add docker && hflow run /wfdir"

Where

• hyperflowwms/soykb-worker is the name of the workflow worker container (Soykb in this case)
• current directory contains workflow.json
• subdirectory inputs contains workflow input data

Outputs:

• Directory inputs will contain files generated by the workflow run
• Directory inputs/logs-hf will contain logs of all workflow jobs

Running in a Kubernetes cluster

See HyperFlow Kubernetes deployment project for more information.

Running in a distributed infrastructure using the RabbitMQ executor (not maintained)

• Start the RabbitMQ container: docker run -d --name rabbitmq rabbitmq:3
• Add option -e AMQP_URL=amqp://rabbitmq
• More information in the hyperflow-amqp-executor project
• Warning: currently not maintained and not tested with latest HyperFlow versions

Local configuration files

You can provide workflow configuration through local configuration files:

• workflow.config.json -- main configuration file
• workflow.config.{name}.json -- any number of secondary configuration files

The content from all configuration files will be merged and passed to workflow functions via context.appConfig. For example for files:

workflow.config.json:
{
  "main": "mainValue"
}

workflow.config.foo.json:
{
   "secondary": "secondaryValue"
}

The following will be passed in context.appConfig:

{
  "main": "mainValue",
  "foo": {
     "secondary": "secondaryValue"
  }
}

HyperFlow server

The HyperFlow engine can be started in a server mode using command: hflow start-server

If succesfully started, the server prints its URL:

HyperFlow server started at http://localhost:38775

Workflows can be run through the HyperFlow server as follows:

hflow run --submit=<hyperflow_server_url> <workflow_dir>

Currently <workflow_dir> must be a local directory accessbile by the server. This allows running multiple workflows (concurrently) using the same instance of the HyperFlow engine.