Skip to content

A Go package to help write migrations with go-pg/pg.


Notifications You must be signed in to change notification settings


Repository files navigation


Version GoDoc Build Status Coverage Status Go Report Card

A Go package to help write migrations with go-pg/pg.



Because go-pg now has Go modules support, go-pg-migrations also has modules support; it currently depends on v10 of go-pg. To install it, use the following command in a project with a go.mod:

$ go get

If you are not yet using Go modules, you can still use v1 of this package.


To see how this package is intended to be used, you can look at the example directory. All you need to do is have a main package (e.g. example); call migrations.Run with the directory you want the migration files to be saved in (which will be the same directory of the main package, e.g. example), an instance of *pg.DB, and os.Args; and log any potential errors that could be returned.

Once this has been set up, then you can use the create, migrate, rollback, help commands like so:

$ go run example/*.go create create_users_table
Creating example/20180812001528_create_users_table.go...

$ go run example/*.go migrate
Running batch 1 with 1 migration(s)...
Finished running "20180812001528_create_users_table"

$ go run example/*.go rollback
Rolling back batch 1 with 1 migration(s)...
Finished rolling back "20180812001528_create_users_table"

$ go run example/*.go help
  go run example/*.go [command]

  create   - create a new migration in example with the provided name
  migrate  - run any migrations that haven't been run yet
  rollback - roll back the previous run batch of migrations
  help     - print this help text

  go run example/*.go create create_users_table
  go run example/*.go migrate
  go run example/*.go rollback
  go run example/*.go help

While this works when you have the Go toolchain installed, there might be a scenario where you have to run migrations and you don't have the toolchain available (e.g. in a scratch or alpine Docker image deployed to production). In that case, you should compile another binary (in addition to your actual application) and copy it into the final image. This will include all of your migrations and allow you to run it by overriding the command when running the Docker container.

This would look something like this:

# Dockerfile
FROM golang:1.13.3 as build


COPY go.mod go.mod
COPY go.sum go.sum
RUN go mod download

COPY . .

RUN CGO_ENABLED=0 GOOS=linux go build -installsuffix cgo -ldflags '-w -s' -o ./bin/serve ./cmd/serve
RUN CGO_ENABLED=0 GOOS=linux go build -installsuffix cgo -ldflags '-w -s' -o ./bin/migrations ./cmd/migrations

FROM alpine:3.8

RUN apk --no-cache add ca-certificates
COPY --from=build /app/bin /bin

CMD ["serve"]
$ docker build -t service:latest .
$ docker run --rm service:latest migrations migrate


While go-pg has its own migrations package, it leaves a bit to be desired. Some additional features that this package supports:

  • Complete migration diffing to determine which migrations still need to be run. Since go-pg/migrations checks the current version of migrations and runs any migrations after that, some migrations can be missed, especially when several people are working on the same project.
  • Timestamp-based prefixes to prevent two people creating a migration with the same version on two separate branches. If the current version is 3, and more than one person branches off and creates a new migration, all of them will be version 4.
  • The ability to run migrations in a transaction on a case-by-case basis. Most of the time, running migrations within a transaction is desirable, so that if it errs out within the "up" function, the whole migration is reverted. But since some long-running migrations might have a statement with a relatively exclusive lock, you might opt out of running that specific migration within a transaction.
  • A migration locking mechanism. This is to avoid two people (or an automated deployment system) attempting to run migrations at the same time against the same database, which could lead to undesired behavior.
  • An expected workflow of how this package should be used within a project. While go-pg/migrations has some recommendations and examples, this package takes a more opinionated approach which makes it so you don't have to think about it as much, and there's less code for you to write and maintain.
  • Batch-level rollbacks. When there are multiple migration files run during the same migration invocation, they are all grouped together into a "batch". During rollbacks, each batch gets rolled back together. This tends to be more desireable since this usually means the application is reverting back to a previous release, so the database should be in the state expected for that release.

Many of these features and expected behaviors come from using Knex.js migrations in production for many years. This project is heavily inspired by Knex to provide a robust and safe migration experience.

go-pg is a great and performant project, and hopefully, this makes it a little better.


To develop on this project, you'll need to have Postgres running because the tests depend on it.

If you have it running on your machine because it was installed through your package manager (like brew or apt-get), you just need to run the following to get it set up correctly:

make setup

If you don't have it on your laptop, you can run the following to start it within Docker:

make postgres

That should start the container and keep it running while you develop. Once you're done, you can ^C out of it and it will stop the container.

To run the tests, you should run:

make test

To run the linter, you should run:

make lint