Workflow with asynchronous action never stops after another step failure

[laurentganne]

Bug Report

Description

A Workflow is running at some point two steps in parallel:

• a standard lifecycle "create" operation on a component
• a job lifecyle asynchronous operation "run" on a job.

The "create" operation fails. Then the corresponding step appears as failed in the workflow, as expected.
But the "run" step of the job component will always appear as running, and no new action to run again the run action is scheduled.
The workflow is seen as running, forever.

To workaround this issue, the administrator will have to change the status of the yorc task manually.
Before this manual change, the internal statuses for tasks and workflow in consul show that:

• the task status is 1 (RUNNING),
• the workflow step corresponding to the create operation of the standard lifecycle component is ERROR
• the workflow step corresponding to the run action of the job component is RUNNING.
• there is no scheduling action registered (as it has been unregistered by yorc because of the other step that failed in the workflow), but there still one single entry _yorc/scheduling/actions/<id>/data/<key> in consul corresponding to an action data that the run operation has updated while the corresponding action had already been unregistered by yorc.

Priority

Medium

• High = We should stop anything we are doing and do this. If you're writing this calmly, you should probably choose another option. If it's really that urgent, please make sure we read this.
• Medium = It's important that we do this within a few days.
• Low = We will consider this on our next sprint planning.

(Please be aware that your priority may not match ours, we'll use this as guidance only).