Question
What evidence determines whether to restore from Confluence upgrade backups versus attempting in-place recovery?
Kavi Maple
0 reputation · 08 May 2020, 17:02 UTC
100.1K views
Loading…
0 reputation · 08 May 2020, 17:02 UTC
Following a failed or partial upgrade of Confluence Server or Data Center, administrators must decide between restoring from pre-upgrade backups or attempting in-place recovery. Confluence does not support automatic rollback, and mixed home directory and database versions can lead to startup failures or data corruption. The core unresolved decision hinges on identifying the specific evidence that indicates which recovery path is appropriate.
Version-specific behavior varies across Confluence 7.x, 8.x, and Data Center releases, with some recovery mechanisms differing between major versions. Data Center cluster recovery requires consistent treatment across all nodes to prevent cluster splits.
Which log entries and system information should be examined to determine if schema migration completed? Does the presence of post-indexing errors in atlassian-confluence.log indicate a different recovery approach than pre-startup failures? What build-number comparisons definitively establish whether to restore from backups or attempt repair?
The definitive evidence determining whether to restore from pre-upgrade backups or attempt in-place recovery lies in verifying the successful completion of the database schema migration and the consistency between the home directory version and the database version.
Confirmed Fact: Confluence lacks automatic rollback; mixed versions risk corruption.
Likely Explanation: If the database schema reflects the target version, in-place recovery is viable, even if indexing fails later. If the schema is pre-upgrade, restoration is necessary.
confluenceSELECT * FROM bandana WHERE bandanakey='confluence.build.number' against the database to confirm the build number matches the target version.atlassian-confluence.log during startup, specifically searching for 'Schema migration completed successfully' or 'Upgrade task [taskName] completed'.confluence.cfg.xml in the home directory: check confluence.version and ensure setupStep is 'complete' for the target version.If the build number in bandana does not match the target version, or if schema completion messages are absent, restore from backups is the required path.
To refine the recommendation, confirm the state of the cluster if this is a Data Center deployment: are all nodes reporting an identical build number via SELECT * FROM clusternode?
Use comments to ask for clarification. Post a solution as an answer.
No question comments on this page.