Editing workflow breaks extra_vars

[blindrood]

ISSUE TYPE

• Bug Report

SUMMARY

Editing workflow generates lots of issues with extra vars on job templates

ENVIRONMENT

• AWX version:6.0.0
• AWX install method: docker
• Ansible version: 2.8.1
• Operating System: Win
• Web Browser: FF/Chrome

STEPS TO REPRODUCE

Edit existing workflow template.

EXPECTED RESULTS

Everything is working as on v3

ACTUAL RESULTS

Workflow designer to some jobs adds pipe to extra vars, like this:

|
  {}

Causing error: Failed to parse variables. Attempted to parse yaml. Parser did not return an object.

On other jobs it splits extra vars content char by char generating list of numbered, one char vars, like this:
https://gist.github.com/blindrood/156a73d31654114508bff7307ef2fdd1

ADDITIONAL INFORMATION

[ghjm]

Possible duplicate of #4303

[mabashian]

@blindrood you'll have to provide some more information to help us reproduce this issue. Can you describe the setup of the job template being referenced in the workflow? What fields are promptable? Does it have a survey? If possible can you provide a gif or screenshot of the error being reproduced?

[blindrood]

Some templates have survey, some doesn't. Seems like, if there's no extra_var then a pipe is added. If there's something, then it's splitted

[blindrood]

We've upgraded awx to 6.1. Extra pipe no longer appears.
Extra variables are still splitted char by char

[kfattig]

We're also seeing this. I'm recording here since there was already a public issue, but we do have Tower support. If you'd like me to file a vendor ticket LMK.

When creating the 'node' in the workflow, things are normal:

The confirmation screen is also normal:

After hitting 'Save', it seems that the extra_variables are written incorrectly to the workflow_node's entry in the database:

select extra_data from main_workflowjobnode where id=31770;

Returns:

{"0": "j", "1": "t", "2": "_", "3": "v", "4": "a", "5": "r", "6": "1", "7": ":", "8": " ", "9": "j", "10": "t", "11": "_", "12": "v", "13": "a", "14": "l", "15": "1", "16": "\n", "17": "j", "18": "t", "19": "_", "20": "v", "21": "a", "22": "r", "23": "2", "24": ":", "25": " ", "26": "j", "27": "t", "28": "_", "29": "v", "30": "a", "31": "l", "32": "2"}

If you re-open the workflow visualizer, each character in extra_vars has become its own var:

They are then passed onto the job at exeuction:

select extra_vars from main_job where unifiedjob_ptr_id=161395; 

Returns:

{"jt_var1": "jt_val1", "jt_var2": "jt_val2", "0": "j", "1": "t", "2": "_", "3": "v", "4": "a", "5": "r", "6": "1", "7": ":", "8": " ", "9": "j", "10": "t", "11": "_", "12": "v", "13": "a", "14": "l", "15": "1", "16": "\n", "17": "j", "18": "t", "19": "_", "20": "v", "21": "a", "22": "r", "23": "2", "24": ":", "25": " ", "26": "j", "27": "t", "28": "_", "29": "v", "30": "a", "31": "l", "32": "2", "wf_var1": "wf_val1", "wf_var2": "wf_val2"}

[mabashian]

Pretty sure this is related to #3669 #3567 #3359

[dsesami]

Manipulated a workflow with both JSON and YAML formatted extra vars. Created and edited the workflow with different variables. There does not appear to be any strange output by the application anymore. 👍