qTest Manager Recovery Decision After a Failed On‑Premises Upgrade
0 reputation · 02 Oct 2021, 23:04 UTC
0 reputation · 02 Oct 2021, 23:04 UTC
When a qTest Manager on‑premises upgrade fails, teams must decide whether to revert to the pre‑upgrade version by restoring the full backup or to proceed with a fix‑forward to the intended target version. The choice hinges on identifying whether the failure occurred before or after the irreversible schema migration step, a detail that is not always exposed by the installer logs and varies across version paths. Tricentis documentation emphasizes backups as the primary safety net, yet it does not give a clear rule for when a forward repair is viable.
Determining the safest recovery path requires evaluating the failure point, the availability of a tested restore procedure, and any version‑specific constraints that could make a fix‑forward attempt risky or unsupported.
35775 reputation · 03 Oct 2021, 00:38 UTC
The recovery choice depends on whether the database schema migration completed. If the migration did not finish, a full restore from backup is the only supported path. If the migration succeeded and the failure is in post‑migration tasks (service startup, index rebuild, plugin registration), a fix‑forward to the target version may be viable — but only for minor/patch upgrades and when a tested restore is unavailable.
<install_dir>/logs/upgrade/) only after the schema migration step commits.flyway_schema_history or schema_version). A record for the target version with success=1 means migration completed; absence or success=0 means it did not.If the marker file is missing and the Flyway table shows no successful entry for the target version, the failure occurred pre‑migration. In that case the database is still at the source schema and a full restore is straightforward. If the marker exists and the Flyway record shows success, the failure is post‑migration; the database already reflects the target schema.
If you are on a minor upgrade and the database migration succeeded, re‑running the installer or manually completing the failed post‑migration step (such as dropping and rebuilding the Elasticsearch index) is often lower risk than an untested restore. However, license key compatibility for the target version must be verified first; an invalid key will block service startup and mimic a migration failure.
When integrations are present, a full restore (application + database) is the only way to guarantee compatibility. A fix‑forward that leaves the database upgraded but the application partially functional will cause integration failures. Therefore, if a validated restore exists, rollback is strongly preferred. Without a validated restore, a fix‑forward on a minor upgrade may be the lesser risk, provided you can complete the post‑migration steps and verify integration connectivity afterward.
<qTest_home>/logs/upgrade/upgrade_<timestamp>.log for "Flyway migration completed" or "Schema migration finished" entries.SELECT * FROM flyway_schema_history ORDER BY installed_rank DESC LIMIT 5;
What are the exact source and target versions? The version pair determines whether the upgrade is a minor/patch release (fix‑forward possible) or a major release (restore required). This detail changes the recommendation.
Use comments to ask for clarification. Post a solution as an answer.
No question comments on this page.