Invoking workflows
After successfully registering a workflow, the workflow is ready to be invoked. This is implemented by the GitHub Action FAASR INVOKE, also available in your FaaSr-workflow repo once you fork it. In short, FAASR INVOKE "kick-starts" your workflow by invoking its entry point action; subsequent actions are then executed according to the order implied by your workflow DAG as per the programming model.
Requirements
To invoke a workflow, you need:
- Your forked FaaSr-workflow repo
- A JSON workflow configuration file, downloaded from the workflow Web UI and uploaded to your FaaSr-workflow repo, successfully registered with
FAASR REGISTER - Proper cloud credentials for any cloud compute servers and data stores you use, stored as GitHub Secrets. These secrets must also be copied and available in cloud platforms that have native secret stores, AWS and GCP
Running FAASR INVOKE
- The workflow invocation action
FAASR INVOKEis available under theActionstab in your FaaSr-workflow repo (it is named such that it appears near the top of the action list). - To execute, select this action from the list, then click on the drop-down
Run workflowmenu to the right, and enter the name of the JSON workflow configuration file. - If invocation is successful, each action in your workflow results in the execution of a a corresponding action in the cloud provider you use: these could be GitHub actions (for GH compute servers), AWS Lambdas (for AWS), Google Cloud Run jobs (for GCP), OpenWhisk actions (for OW), or Slurm jobs.
- The naming scheme used for the actions invoked throughout the workflow by
FAASR INVOKEis as follows:WorkflowName-ActionNamewhereWorkflowNameis the workflow name defined in your configuration file (underWorkflow settingsin the workflow Web UI), andActionNameis the name of the action (each node of the graph in the workflow Web UI). - For example, for the FaaSr tutorial, the workflow name is
tutorialand there are two actions in the graph (startandsum); correspondingly, two GitHub Actions are executed, in sequence, namely:tutorial-startfollowed bytutorial-sum. - If any errors occur during invocation, a red
Xicon will appear after the action completes. Click on the action name and then on thedeploybox to review logs for errors. - Errors that may happen during the workflow execution will appear with the same red
Xicon, if you use GH as your cloud compute platform. If you invoke workflows in other platforms (AWS, GCP, OW, SLURM), error logs will be available on that cloud platform rather than on GitHub Actions.