Recently we have been migrating our Connect apps to Forge; and as part of this process we have needed to use forge version bulk-upgrade to move customers from our Connect version to our Forge version.
When we did this for our first app (with the fewest number of total installations), every installation upgraded successfully. Running forge version bulk-upgrade list reported “0 pending” and “0 failed” installations, including both active and inactive license states.
For our second app, the vast majority of installations upgraded successfully, but not all. Running forge version bulk-upgrade list reports “4 failed”. Repeating the bulk upgrade multiple times has not been successful in getting these last 4 installs upgraded.
All four of those failures are for installations where the license status is inactive (yet many other installs with inactive licenses did upgrade successfully).
The fact that these are all inactive gives us some comfort, however as we’re looking to switch off our old Connect infrastructure we are concerned about these installs that remain on our Connect version.
I realise this might be a difficult question to answer in a general sense, but what would be some reasons why an upgrade fails; and is there any way to troubleshoot failures when using the bulk-upgrade tool?
This is most often when the customers workspace has been shut down and therefore, since Connect is intimately tied into JIRA and Confluence (in a way that Forge is not), the migration from Connect to Forge (or even Forge with Connect modules to Forge without) can’t be safely performed.
If this is critical to you, I would recommend that you react out to the support team and they can give you exact failure reasons to confirm.
Otherwise you can wait a month or two for those customers to reactivate / permanently delete.
Sorry, this is a pain point and something that we hope to schedule in the future but right now are prioritising unblocking the select remaining forge feature requests needed for all apps to move to Forge.
@scottohara I’ve been liking your old posts recently because I’m having the same issues. So they haven’t fixed much since then. Migration process is still a broken, opaque and undocumented mess.
- The automated private version bump problem still exists and is undocumented.
- Automated migration of Connect installs simply does not work for me at this point in time, and has no insight into the status other than waiting 96 hours and praying you still have paying customers at the 97th hour.
- Connect infrastructure retirement step continues to have absolutely zero documentation or guidance. Surely this is a critical piece of the migration process?
Now I’m wasting time reaching out to the key decision makers at Atlassian to fix these basic reported gaps in the migration process that should have been addressed years ago.
I’m only migrating now because they’re threatening to plaster end-user app iframes with warning messages sometime either in “late September” or “eventually”.
Gonna tag @rmassaioli just so he sees this too.