[ML] Improve copy config to json experience (export/import of ML adv configuration)

[elasticmachine]

Original comment by @ppf2:

5.5

Sometimes users will want to duplicate ML configuration to another environment by copying the raw json (esp. for an advanced configuration).

Currently, if they do this from an existing configuration's JSON tab and use the copy button:

When importing, it will throw an unintuitive error:

Save failed: [illegal_argument_exception] [job_details] unknown field [data_counts], parser not found

This simply means that the json being copied has sections in them that have already been enriched to include previous run metrics so that if you now use this copied json in a new adv config, it will not work.

The workaround is to use clone job, and then copy the json generated there (intermediate) in the cloned job (but not actually cloning the job itself). This will produce a valid json and the only thing they have to manually edit is to fill in a job id, etc..

Perhaps we can consider providing 2 copy to json buttons from the existing job screen, 1 for copying the json with previous run metrics, 1 without. A better solution will be: When someone tries to save an "exported" json in a new advanced configuration, parse that config, if it sees sections that have previously run metrics, automatically clean those up so that the json will then be a valid configuration to use in a new adv config.

[elasticmachine]

Original comment by @richcollier:

bumping this issue - customers are now starting to ask to migrate ML jobs from one cluster to another (i.e. dev to production).

It would be nice to have a concise method to move both job/datafeed configs as well as model/results from one cluster to another

[elasticmachine]

Original comment by @cwurm:

A customer ran into this issue in their ML evaluation. I think the intuitive thing to do would be to either:

1. Not have run metrics in the JSON in the UI
2. Not exporting them with the copy JSON button (this might be weird since it's visible but wouldn't be in the export)
3. Ignore the run metrics when pasting a JSON in the Advanced Job config UI -- as Pius suggested.

[elasticmachine]

Original comment by @sophiec20:

Regarding 1 and 2 above (removing run metrics) - just highlighting that it can be quite useful to have the run metrics in the JSON for the scenario where this gets sent to us as a support ticket.

[sophiec20]

Closing this as out of date. The advanced wizard was refreshed in 7.5 and "copy to clipboard" no longer exists.

Related:

• From 7.11 (the next release), at the API level it is possible to define query parameter ?exclude_generated which will exclude the run metrics from the output of GET job making that output more suited for re-use.
• Ability to move anomaly jobs already exists as a github ticket [ML] Ability to move anomaly detection jobs between clusters elasticsearch#37987