A green migration test only proves that the migration path it actually ran succeeded against the database state the job received. If a previous job left behind a database that was already migrated—or whose schema or migration history had changed—the next job may skip the transition you intended to test. Check whether CI reuses the database, inspect both its schema and migration history, then rerun the test against a newly created database in a known starting state.
Why can a leftover database make a migration test pass?
A migration test is meaningful only relative to its starting state. If the test expects an older schema but receives a database already at the newer migration state, the migration may have nothing to do. The command can exit successfully without proving that the change works from the intended baseline.
As an Amazon Associate I earn from qualifying purchases.
That can happen when CI shares a database between jobs, restores a database volume from a cache, or preserves a test database after a run. The exact cause depends on the project’s framework and CI configuration; the title alone does not identify either.
Database lifecycle behavior is not universal. Django’s test runner normally destroys test databases after a run, but --keepdb opts into reuse, and a forced interruption can leave a test database behind. Check the project’s actual command and cleanup path rather than assuming a successful or failed job removed its database. Django documents its test database lifecycle.
#1 Best Overall
How do you confirm whether the job reused stale state?
- Trace how CI obtains the database. Determine whether each job provisions a new database, connects to a shared one, or restores a cached or persisted volume. Review setup and teardown scripts, framework flags, and cancellation or timeout handling.
- Capture the starting state before migrations run. Record the schema and the migration-history records, then compare both with the baseline the test is supposed to use. An exit code alone does not reveal which starting state the job received.
- Check migration bookkeeping and schema separately. For EF Core, pending migrations are determined by comparing migrations in the application assembly with records in the target database. That comparison does not prove that the database’s actual schema matches those records or that it began at the intended baseline. Inspect both. EF Core explains applied and pending migration management.
- Repeat against a clean, disposable database. Create or reset a throwaway database to the expected starting state, apply migrations, and verify the resulting schema and any important data transformations. Supabase’s documented workflow, for example, resets the local database before applying migrations. Supabase’s migration guide shows that workflow.
What should a trustworthy migration check verify?
A reliable check validates the transition, not merely that a migration command returned successfully. It should establish the starting point, exercise the intended migration path, and verify the result.
- Starting state: Confirm the database is new or reset to the expected baseline, not merely present and reachable.
- Migration execution: Confirm the migrations that should run actually ran. A command that sees no pending work is not evidence that the intended transition was tested.
- Resulting state: Assert expected schema changes and data transformations. Migration-history metadata and the physical schema are related but distinct things to validate.
- Repeatability: Make cleanup explicit for success, failure, timeout, and cancellation so a later job cannot silently inherit state.
These checks catch different failure modes: bookkeeping can say a migration was applied when the schema is wrong, or schema changes can exist without matching migration records. GitLab documents a migration-check job that compares schema after rollback and checks generated versus committed migration history. See GitLab’s migration-check job.
How should CI handle cleanup and concurrent tests?
Prefer an isolated database for each job when practical. If a shared database is necessary, make ownership, reset behavior, and serialization explicit; otherwise, one job can alter the state another job expects.
Cleanup must also account for tests that manage transactions. EF Core describes rollback-based isolation for many database tests, but tests that manage transactions need appropriate cleanup, and tests against the same shared database should not run in parallel when that can cause interference. EF Core’s database-testing guidance covers these isolation considerations.
Rank #3
- Dell PowerEdge R730xd 24B SFF 2U Server
- 2x Intel Xeon E5-2690 v4 2.6Ghz 14-Core (28-cores Total)
- 128GB DDR4 RAM – 4x 1.2TB 10K SAS 2.5” 12Gb/s
- Dell H730P mini 2GB 12Gb/s RAID
- 2x 750W PSU - 2x 10Gb SFP+ 2x 1Gb (RJ45) NIC
What if the migration already ran on a shared database?
Do not delete migration source code as a shortcut if that migration has already been applied to shared databases. Keep the applied migration available and make a corrective migration, or coordinate a rollback that accounts for every database using it. Before applying changes to production, inspect the generated SQL script and confirm it is appropriate for the database’s current migration state: EF Core notes that scripts apply only from the appropriate migration state. EF Core’s guidance on applying migrations explains script use, while its migration-management guidance addresses managing migration history.
Quick Recap
Rank #4
- Server 2022 Standard 16 Core
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
