5.4 KiB
forgejo_migrations
Forgejo has three database migration mechanisms:
models/gitea_migrations
- Original database schema migrations from Forgejo's legacy as a fork of Gitea.
- A linear set of migrations referenced in
models/gitea_migrations/migrations.go
, each represented by a number (eg. migration 304). - The current version is recorded in the database in the table
version
.
models/forgejo_migrations_legacy
- The next 50-ish database schema migrations reflecting change in Forgejo's database structure between Forgejo v7.0 and v14.0
- A linear set of migrations referenced in
models/forgejo_migrations_legacy/migrate.go
, each represented by a number (eg. migration 43). - The current version is recorded in the database in the table
forgejo_version
.
models/forgejo_migrations
- The most recent database schema migrations, reflecting change in the v14.0 release cycle and onwards into the future.
- Each migration is identified by the filename it is stored in.
- The applied migrations are recorded in the database in the table
forgejo_migration
.
forgejo_migrations
is designed to reduce code conflicts when multiple developers may be making schema migrations in close succession, which it does by avoiding having one code file with a long array of migrations. Instead, each file in models/forgejo_migrations
registers itself as a migration, and its filename indicates the order that migration will be applied.
Files in forgejo_migrations
must:
- Define an
init
function which registers a function to be invoked for the migration. - Follow the naming convention:
- The letter
v
- A number, representing the development cycle that the migration was created in
- A letter, indicating any required migration ordering
- The character
_
(underscore) - A short descriptive identifier for the migration
- The letter
For example, valid migration file names would look like this:
v14a_add-threaded-comments.go
v14a_add-federated-emojis.go
v14b_fix-threaded-comments-index.go
Migration Ordering
Forgejo executes registered migrations in forgejo_migrations
in the strings.Compare()
ordering of their filename.
There are edge cases where migrations may not be executed in this exact order:
- If a schema change is backported to an earlier Forgejo release. For example, if a bugfix during the v15 development cycle was backported into a v14 patch release, then a migration labeled
v15a_fix-unusual-data-corruption.go
could be applied during a v14 software upgrade. In the future when a v15 software release occurs, that migration will be identified as already applied and will be skipped. - If a developer working on Forgejo switches between different branches with different schema migrations.
- If the contents of the
forgejo_migrations
database table are changed outside of Forgejo modifying it.
Creating a new Migration
First, determine the filename for your migration. In general, you create a new migration by starting a file with the same prefix as the most recent migration present. If v14a_add-forgejo-migrations-table.go
was the last file, most of the time you can create your migration with the same v14a_...
prefix.
There are two exceptions:
- After the release branch is cut for a release, increment the version in the migration. If v14 was cut, you would start
v15a_...
as the next migration. - If your migration requires that an earlier migration is complete first, you would increment the letter in the prefix. If you were modifying the table created by
v14a_add-forgejo-migrations-table.go
, then you would name your migrationv14b_...
.
Once you've determined the migration filename, then you can copy this template into the file:
// Copyright 2025 The Forgejo Authors. All rights reserved.
// SPDX-License-Identifier: GPL-3.0-or-later
package forgejo_migrations
import (
"forgejo.org/modules/timeutil"
"xorm.io/xorm"
)
func init() {
registerMigration(&Migration{
Description: "short sentence describing this migration",
Upgrade: myMigrationFunction, // rename
})
}
func myMigrationFunction(x *xorm.Engine) error {
// add migration logic here
//
// to prevent `make watch` from recording this migration as done when it
// isn't authored yet, returh an error until the implementation is done
return errors.New("not implemented yet")
}
And now it's up to you to write the contents of your migration function.
Development Notes
Once migrations are executed, a record of their execution is stored in the database table forgejo_migration
.
=> SELECT * FROM forgejo_migration;
id | created_unix
-----------------------------------+--------------
v14a_add-forgejo-migrations-table | 1760402451
v14a_example-other-migration | 1760402453
v14b_another-example | 1760402455
v15a_add-something-cool | 1760402456
v15a_another-example-again | 1760402457
If your migration successfully executes once, it will be recorded in this table and it will never execute again, even if you change the migration code. It is common during development to need to re-run a migration, in which case you can delete the record that you're working on developing. The migration will be re-run as soon as the Forgejo server is restarted:
=> DELETE FROM forgejo_migration WHERE id = 'v15a_another-example-again';