Blocked on testing the Connect to Forge upgrade: descriptor-URL install removed from dev sites, and the 30-day custom field retention breaks migratedFromConnect

We are migrating a published Jira Cloud Connect app to Forge as an in-place upgrade, keeping the same app key so existing customers get Forge as an app update.

Two things block us from validating that upgrade before it reaches customers.

1. No way to point a dev site at a non-production backend

Our Connect descriptor baseUrl decides which backend the app talks to. We used to install the Connect app on dev sites from a descriptor URL under Development Mode, which pointed each site at a dev backend. The Connected apps screen now says:

Development Mode no longer supports installing private apps using an app descriptor URL.

The only route left is Marketplace, whose descriptor points at our production backend, so a dev install would send test traffic to production. Existing dev installs are stuck on whichever backend they were created against.

Is there a supported way to install a private Connect app on a dev site by descriptor URL, or to override the baseUrl of an existing Connect installation on a specific site?

2. The 30-day custom field retention blocks migratedFromConnect

Our Connect app owns a custom field, so the Forge app declares the equivalent jira:customField with migratedFromConnect set to true, per Migrating Jira issue fields.
That page’s Limitations section says:

The migration task will fail to switch a custom field from Connect to Forge if a custom field with the same key was already installed.

The Field lifecycle section of the jira:customField module reference says an uninstalled app’s field is retained for 30 days and “will reappear with all the previous data intact” on reinstall, and that the data is permanently lost only after that.

Measured on one of our dev sites, a field still inside that window is enough to trip the Limitations precondition, and the refusal is silent and destructive:

  • The Connect field was deleted, along with its value on every issue.
  • A new, empty Forge field was created at a new customfield id, instead of the identifier being preserved as the docs describe for a successful switch.
  • The Task progress API returned “No ongoing migration task for the custom field”, which is indistinguishable from a completed run. No error appeared anywhere.

We also could not detect the retained field beforehand. In that window it does not appear in
/rest/api/2/field, /rest/api/3/field/search, /rest/api/3/field/search/trashed, or the admin Deleted fields screen, yet a reinstall restores it with its data. So a site cannot be confirmed clean by inspection.

Is there a supported way to shorten the 30-day retention on a dev site, to fully delete a Forge custom field and its retained record after uninstall, or to have Atlassian clear it for a named dev site?

Why this blocks us

We have a small, fixed set of dev sites carrying the Connect app, and without descriptor-URL install that set cannot be renewed. Any site that has had the Forge app installed is inside the retention window and cannot validate the migration. Any site that has not is a single attempt, because a refusal destroys the Connect field’s values and the Connect app cannot be reinstalled to recreate them.

So we are preparing to ship a one-way, irreversible field migration to production customers without having seen it succeed anywhere. We would rather test it properly first.

Happy to share app keys, site identifiers and timestamps privately if that helps someone look at a specific case.

Thanks!!

Hi @RohitSinghavi,

Thanks for the detailed write-up.

Since this involves specific sites, app keys and field data, please raise an ECOHELP ticket through developer support and include a link to this post. We’ll work through a test approach with you there.

Cheers,

Mitch

Thanks Mitch.

We already have an open ticket covering this work, so rather than raising a new one I have added a link to this post there:

The two things we most need out of it are a dev site with no retained custom field record, and a way to point a dev site’s Connect install at a non-production backend. Happy to take either or both on the ticket.

Thanks again.

@MitchMcCue is there any way to escalate the Echohelp ticket or increase its priority so we can get support sooner? We haven’t received any response for over two weeks, and this delay is impacting our testing and release plans.

We’d really appreciate it if you could help us expedite the support.

TIA!!