Migrating Jira and Confluence from Data Center to Cloud (1,000+ projects and spaces), we hit 15 issues that needed manual fixes, mostly broken macros, links and filter IDs, plus Xray and Forge app gaps. Each issue has the symptom, cause, fix and ticket, and there’s a pre-migration checklist at the end. Several are still open, so I’d love to hear how you solved them.
Scale and context: Jira DC 9.5+, Confluence DC 8.5+, 1,000+ projects and spaces combined, migrated with JCMA/CCMA [1.12.53 higher/3.13.6 higher].
Caveat: Everything here is what we observed in our environment, at the time of our migration, with the Migration Assistants and Cloud behaviour as they were then. Ticket statuses change, and some of these may have been fixed since. Please check the current status before relying on any of it.
What went well: Core content, users and most Jira data migrated cleanly, and Atlassian Support was responsive on every case below. The issues here are the edge cases worth planning for.
At a glance
Impact: High = broke content or blocked users; Medium = needed rework or caused confusion; Low = cosmetic or has a workaround.
# | Product | Issue | Impact | Related ticket |
|---|
1 | Confluence | External links to DC pages stop resolving | High | MIG-507 |
2 | Confluence + Jira | Jira Macro Repair left some macros unmapped | High | |
3 | Confluence | Jira macro count in Macro Usage differs from DC | Medium | — |
4 | Confluence | Macros inside comments show as greyed out/unsupported | High | CONFCLOUD-74921 |
5 | Confluence (Jira macros) | filter = <id> JQL returns wrong or no results | High | CONFCLOUD-75825 |
6 | Confluence (Jira Chart) | Chart macros error after migration | High | CONFCLOUD-84160 |
7 | Confluence | Children Display macro loses sort order below level 1 | Low–Medium | — |
8 | Jira | Teams duplicated after multiple migration runs | Medium | CONFCLOUD-84033 |
9 | Jira | Pre-flight / data-preparation checks not visible in the UI | Medium | — |
10 | Jira + Confluence | No pause/resume for a running migration | Medium | — |
11 | Confluence | DC became unstable during pre-migration checks (120-space batch) | High | MIG-1961 |
12 | Confluence | Missing attachments flagged by CCMA checks | Medium | CONFSERVER-41622 |
13 | Confluence + Xray | Xray test result gadget unavailable | High | XRAYCLOUD-2393 |
14 | Jira + Xray | Wrong counts in Xray “Tests by Test Type” gadget | Medium | XRAYCLOUD-10728 |
15 | Confluence | Slow page loads with many Forge-app macros | Medium | ECO-1088, CONFCLOUD-82734, FRGE-1495 |
Confluence: links and macros
1. External links to DC pages no longer resolve
What we saw: Confluence pages were linked from several external applications. After migration those links stopped working, because page IDs and URLs differ in Cloud.
What we did: Working with Atlassian Support, we obtained a SQL query that produced a mapping of old DC page URLs to new Cloud URLs, then used the mapping files to update the external references. The approach depends on your environment, so validate the method with Support before running anything against your database.
Ticket: MIG-507, “Provide the corresponding URL information between Confluence Server and Confluence Cloud after migration” (check current status). Atlassian-hosted automatic redirection is tracked separately in MIG-2363.
Suggestion: Inventory every external system that links into Confluence before migration.
2. Jira Macro Repair did not map every macro
What we saw: After running Jira Macro Repair, some Jira macros were still not mapped to the new Jira Cloud URLs.
What Support found: Our Confluence had two Jira application links configured (a primary Jira and a system Jira), and many macros had been created against different server IDs. Confluence relies on the application links to resolve the source Jira URL for each macro, so Support investigated where that URL was stored in the macro data.
What we did: Support provided a SQL query. We exported the results to a .csv of the unmapped macros and their target URLs, and used it to complete the mapping.
Suggestion: Review your Application Links and the server IDs your macros reference before you run Macro Repair.
3. Jira macro count in Macro Usage differs from DC
What we saw: After Macro Repair, the count in the Macro Usage section did not match DC.
What Support told us: In our case, two factors affected the count: backend indexing, and permissions or restrictions on the Jira links.
What we did: We used the REST API to get the count instead of relying on the UI figure. [add the endpoint you used]
Suggestion: Validate macro counts through the API, not the UI.
4. Macros inside comments appear greyed out
What we saw: Some DC pages had macros embedded in comments. In Cloud, those comments showed as greyed out in the editor.
Impact: There was meaningful content loss. The only option we found was to recreate the affected pieces manually after migration example for the Drawio diagrams being embedded.
Ticket: CONFCLOUD-74921, “Improve handling of unsupported macros added to comments”. Atlassian confirmed it on that ticket.
Suggestion: Review macros inside comments before migration and separate them by content type. In our case, Requirement Yogi links and keys embedded in comments could be converted to a text-based format using Groovy scripting, which helped preserve the reference information when the original macro became unsupported in Cloud. However, this approach was not suitable for visual macros such as Draw.io diagrams. For those, we identified the affected content and recreated the macros manually in Cloud.
5. Saved filters referenced by ID return errors
What we saw: Jira macros on Confluence pages that referenced a saved filter by ID (filter = <id>) broke after migration, because filter IDs are different in Cloud.
What we did: We obtained the old-to-new filter ID mapping from the Atlassian migration team and updated the macros manually. The IDs change because Cloud’s underlying data model differs from DC.
Ticket: CONFCLOUD-75825, “JQL using filter id does not show expected results after migration.” This was open when we raised it; check whether it has since been resolved, and which Migration Assistant version you’re on. [
6. Jira Chart macros error after migration
What we saw: Jira Issues macros were repaired, but Jira Chart macros were not. Charts that used a custom field on the X-axis kept pointing to the old field ID, which no longer exists in Cloud, so they failed to render.
Why Macro Repair missed them: The two macros store their query differently, so the repair doesn’t touch the chart.
Macro | Query property |
|---|
Jira Issues | jql-query |
Jira Chart | jql |
What we did: We fixed the affected charts by hand.
Ticket: CONFCLOUD-84160, “Jira Chart macros shows execution error after migration”.
7. Children Display macro loses ordering below level 1
What we saw: A one-level Children Display macro kept its order. With two or more levels, the order was lost.
What Atlassian said: Confirmed as a bug.
Workaround: We changed the sort order to “current.” Sorting still doesn’t apply to level 2 and deeper child macros.
Jira
8. Teams duplicated after multiple migration runs
What we saw: Teams appeared twice in Jira Cloud, and users were unsure which to use.
Why: Unlike DC, Teams are tied to the product rather than a project, so each migration run (test sandbox or production) can create its own copy.
What we did: We listed the team IDs through the REST API and deleted one of each duplicate.
Ticket: CONFCLOUD-84033 (check current status).
Suggestion: After any test migration, clean up its teams before the next run.
9. Pre-flight and data-preparation checks not visible in the UI
What we saw: When running pre-flight checks for Jira projects, the data-preparation checks and the JSM validation did not behave as expected. The data-preparation checks ran for about 20 hours but were not visible in the UI, so we couldn’t track them.
What went wrong: Because we couldn’t see them, we started further checks while earlier ones were still running, which put strain on the database. We also updated, and at one point uninstalled, JCMA while checks were in progress.
Lesson: Don’t start new checks, and don’t update or uninstall JCMA, while checks are running. Confirm with Support whether a check is still active before acting.
Migration tooling and stability
10. No pause/resume for a running migration
What we saw: Our enterprise-scale DC instance was occasionally unstable during the migration, so we asked whether a migration could be paused.
What Atlassian told us: Atlassian confirmed that pause/resume was not available for our migration workflow. We were advised to ensure the source environment was stable before starting the run.
Lesson: Treat source stability as a precondition. Check cluster health first, and schedule runs for low-traffic windows.
11. DC instability during pre-migration checks (120-space batch)
What we saw: Confluence DC went down from high CPU utilization while running pre-migration checks on a batch of 120 spaces. After the service was restored it stayed unstable and went down again. While we were updating a third-party app (Requirement Yogi) to a stable version, DC had further issues. Afterwards, the pre-migration checks stayed stuck “in progress.”
What Support found: There was a stuck thread, and the errors showed the connection between the Migration Assistant and the server timing out (java.net.SocketTimeoutException), so the UI could not communicate with the server.
What we did: Following Support’s guidance: disable JCMA/CCMA, restart the server, and clear the plugin cache. The stuck entry was removed after the restart. Take a backup and follow Support’s steps for your version before doing this.
Ticket: MIG-1961
Lesson: Run checks off-peak, on a stable instance, and don’t change apps or plugin versions mid-check.
12. Missing attachments in CCMA checks
What we saw: After syncing attachments from sandbox to production, CCMA attachment checks reported missing attachments: the file path did not exist on the server.
Cause: The files were genuinely missing from the server itself. Most were .tmp files. Confluence creates temporary files while uploading and processing attachments (for example text extraction and indexing), and our initial analysis suggested a third-party draw.io plugin was adding them.
Options: Resolve the missing attachments before migrating, or, with Support’s guidance, bypass the check by enabling the Migration Assistant dark feature and disabling migration-attachment-check. Bypassing means those files will not be migrated, so confirm that’s acceptable first.
Ticket: CONFSERVER-41622 (check current status).
Marketplace and Forge apps
13. Xray test result gadget unavailable in Confluence Cloud
What we saw: Many DC pages had the Xray Overall Test Result gadget embedded. After migration it was gone.
Cause: A limitation on the Xray vendor side combined with how Jira gadgets work in Cloud.
Ticket: XRAYCLOUD-2393 (check current status).
14. Wrong results in Xray “Tests by Test Type” and invalid test steps
What we saw: After migration, the Test Type dashboard gadget showed inaccurate issue counts, and some test cases showed as failed or had invalid test steps in the new environment.
What we tried: Upgrading Xray and reinstalling the app restored functionality in some cases, but re-indexing kept failing in many projects. The vendor identified a bug.
Ticket: XRAYCLOUD-10728 (check current status).
15. Slow page loads with many Forge-app macros
What we saw: Pages with multiple Forge-app macros loaded slowly, which hurt viewing, editing and updating pages. We have no workaround that fully solves it.
Tickets: ECO-1088, CONFCLOUD-82734, FRGE-1495 (check current status).
Pre-migration checklist
- Use the newest version of JCMA and CCMA.
- Inventory external systems that link into Confluence, and plan URL mapping.
- Review Application Links and server IDs referenced by Jira macros before running Macro Repair.
- Back up macros embedded in comments.
- List saved filters and custom fields used in macros and charts (IDs will change).
- Check Marketplace apps (Xray, draw.io, Forge apps) for Cloud parity and known gaps.