WF execution is not created if input preparation of initial task fails
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mistral |
Fix Released
|
High
|
Nikolay Makhotkin |
Bug Description
If an issue occurs on a stage where we're preparing task input data, for example, a workflow execution will not be created and the execution-create response will be different.
The first task(s) execution objects get created simultaneously with WF execution, within the same transaction
if something is wrong with preparing at least with one of those objects then entire TX rolls back
In case of a WF with two starting tasks in parallel and one is OK and the other throws an error on input preparation the whole WF execution object won't be created.
A successful start task should be saved in DB for sure (as well as WF execution), otherwise we loose all tracking information
This can be reproduced with the following WF:
---
version: '2.0'
fail_on_initial:
type: direct
tasks:
iniitial_task:
action: std.echo output=<% $.non_existing_
description: | updated |
Changed in mistral: | |
milestone: | none → mitaka-1 |
importance: | Undecided → High |
status: | New → Triaged |
Changed in mistral: | |
assignee: | nobody → Nikolay Makhotkin (nmakhotkin) |
tags: | added: liberty-backport-potential |
Changed in mistral: | |
milestone: | mitaka-1 → 2.0.0 |
Fix proposed to branch: master /review. openstack. org/239638
Review: https:/