[CT-1003] run_results.json not generated when on-run-end fails

[seunghanhong]

Is this a new bug in dbt-core?

• I believe this is a new bug in dbt-core
• I have searched the existing issues, and I could not find an existing issue for this bug

Current Behavior

The dbt artifact run_results.json is not generated when a statement in the on-run-end fails. This behavior does not occur when the statement fails in a post-hook.

Expected Behavior

run_results.json must be generated even if a statement fails at the on-run-end and must contain dbt model execution results to indicate which models ran successfully or failed.

Steps To Reproduce

1. Create a sample dbt project with dbt init
2. Run dbt build and confirm target/run_results.json exists.
3. Run dbt clean to remove the target directory.
4. Add on-run-end: "select col1" in the dbt_project.yml and it causes dbt a Database error.
5. Run dbt build and confirm target/run_results.json does not exist.

Relevant log output

No response

Environment

- OS: MacOS 12.3.1
- Python: 3.8.5
- dbt: 1.1 and 1.2

Which database adapter are you using with dbt?

postgres, redshift

Additional Context

No response

[lostmygithubaccount]

I am able to reproduce this easily enough. This should be fixed. Thanks for reporting @seunghanhong!

Acceptance criteria: runs failing from on-run-end hooks should still results in a run_results.json entry with details on the failure.

[iknox-fa]

N.B. Be sure to check with cloud that the existence of this file isn't being used as a indicator of a successful run.

[jtcohen6]

N.B. Be sure to check with cloud that the existence of this file isn't being used as a indicator of a successful run.

My sense is that:

• dbt Cloud job scheduler uses the exit code produced by dbt-core's CLI to determine whether the job succeeded. In this case, an error was raised, so it should be a nonzero exit code (error) regardless
• If run_results.json is not produced, it does prevent the Metadata API from ingesting that artifact and being able to report on that run. Resolution of this ticket will allow the Metadata API to include information about runs with on-run-end hooks that raise errors.