Cloned experiment hits constraint error based on `(branch, experimentId)`.

[ncalexan]

I can't reproduce this -- more on that later.

I cloned https://experimenter.services.mozilla.com/nimbus/set-firefox-as-default-pdf-handler-on-windows-for-new-users-experiment/ from https://experimenter.services.mozilla.com/nimbus/set-firefox-as-default-pdf-handler-on-windows-for-new-users-rollout/. The latter is an 80% rollout with a single "branch", which I named "Treatment" since it has the feature enabled. The former is a 20% side-by-side experiment with two branches, a control and a treatment. If I name the second branch "Treatment", I saw a constraint error, something like (branch, experimentId) is not unique. There was a numerical value for experimentId: 211.

It rather looked like the experimentId of the clone was the same as the original (since the branch names were the same). That seems like a problem.

I mitigated by changing the second branch name to "Enabled". But I can't reproduce by changing the name of the second branch back to "Treatment", suggesting the experimentId has changed, or something else has changed. I don't want to create a second clone to try again, which will pollute the namespace yet further. Sorry for the vague bug report!

┆Issue is synchronized with this Jira Task ┆Attachments: branch rename.mp4

[data-sync-user]

➤ Maria Heres commented:

This issue is also reproducible on Stage for non-cloned experiments. The STR:

1. Have an experiment draft open.
2. Set up a valid branch from “Branches” (i.e. has a feature, name and description).
3. Click “Save”.
4. Rename the branch.
5. Create a new valid branch and give it the initial name of the renamed branch.

The page will display the “duplicate key value violates unique constraint "experiments_nimbusbranch_slug_experiment_id_eefe3985_uniq" DETAIL: Key (slug, experiment_id)=(, ) already exists. ” error message.

Included is a recording of the issue.

!branch rename.mp4|width=1900,height=982!

[jaredlockhart]

This should be fixed now.

[data-sync-user]

➤ Maria Heres commented:

I am not able to reproduce the issue using the STR from my previous comment anymore.