[ML] Job Validation - model memory format check is too generous and is sometimes skipped

[pheyos]

Found in version

• 6.3.0 b1005

Steps to reproduce

• Create a multi metric job (e.g. for the cloudwatch dataset)
• Provide a detector and a valid job name, but no data split
• For Model memory limit set an invalid value (e.g. abc)
• Validate the job => [1]
• Add a data split
• For Model memory limit set an invalid value of the form <number><string> or just number (e.g. 1abc or 5)
• Validate the job => [2]

Expected result

• [1] and [2] : The model memory format check fails and an error message is displayed in the validation results.

Actual result

• [1] the model memory format check is skipped together with the other mml checks. This check should always be executed (as long as all basic checks are successful).
• [2] the model memory check doesn't fail.

Additional information

• [2] the check fails correctly for some other model memory formats, e.g. 0m, abc, 0.

[elasticmachine]

Pinging @elastic/ml-ui

[pheyos]

The issue doesn't occur in 6.5.0 any more. Closing.