Why hasn't Atlassian migrated its own apps?

Forge
1

Quick audit today of the 157 public listings across Atlassian and Atlassian Labs:

  • 36 Forge Cloud
  • 11 Connect Cloud
  • 15 Cloud, neither Forge nor Connect
  • 43 Server-only
  • 11 Data Center-only
  • 41 Server and Data Center

Only 25% of Connect-origin apps appear fully migrated; at least 55% remain on Connect.

I discovered these undocumented migration issues this month (among many others):

  • Marketplace creates a private 2.1.0 while 2.0.0 is under ECOHELP review. Publishing 2.1.0 during review requires resubmission, yet customer installations are auto-migrated to the private version. MP-299 remains unresolved.
  • Documentation previously stated 24 to 96 hours in one place and 120 hours in another. The actual rollout window is up to 10 days and restarts with any new version release.
  • Partners cannot inspect or retry the rollout, must infer its completion from licence data, and have no documented Connect infrastructure-retirement process.

Why are these still occurring and undocumented this late into the migration programme?

Atlassian should have dogfooded the entire migration process across all of its own apps, including Server and DC. That would have surfaced these issues, allowed them to be fixed earlier, and produced accurate documentation for partners.

Meanwhile, the latest changelog confirms that admin warnings are rolling out to Connect apps months before EOS. The previously unannounced iframe-warning plan first appeared in a thread reply, with a deadline of “late September at the latest”; it is now scheduled for “eventually”. EOS has also moved from 31 December 2026 to 31 January 2027.

Will Atlassian’s own Connect apps receive the same warnings?

2

Building my first Forge app rather than migrating one, so I have no Connect scar tissue to trade - still the audit is doing something useful beyond the complaint.

If more than half of Atlassian’s own apps are still on Connect, then Atlassian hasn’t run its own migration end to end. That’s probably not deliberate. The platform team owns the migration and the product teams own their own roadmaps, and nobody inside gets credit for moving an app that already works. So the rough edges stay hidden until a partner hits them.

The dates aren’t the real problem though. The real problem is the one near the end of your post. You start a rollout, you can’t watch it, you can’t retry it, and you end up reading licence data to guess whether it finished. That’s what keeps bringing people back here.

If you push for one thing, push for being able to see the rollout and retry it. The date has already moved once and it can move again, and none of that helps you if you still can’t tell what’s happening.