how to clean up incomplete alembic run

问题

I'm trying to use alembic with a MySQL engine to perform online migrations. I've found that when an operation in my onupgrade() method fails my database gets stuck in an inconsistent state and i can't use alembic until I manually clean up any operations that happened before the failure in onupgrade()

Example:

def upgrade():
    op.create_table('sometable',
            Column('id', INTEGER, primary_key=True),
            Column('name', VARCHAR(150), nullable=False, unique=True))
    op.add_column('anothertable' Column('id', INTEGER))
    op.create_table('secondtable')

So if I run this and the op.add_column fails, even if I fix the add_column line, now "sometable" exists so the first operation will always fail. I can't run my downgrade script, because alembic never updated the version since it didn't complete the upgrade.

I was thinking if there was a way to force run my ondowngrade(), that might be useful. I'd have to ignore errors, as there are sure to be some. Like dropping "secondtable". I couldn't find anyway to do this though.

Anyone have a good way to handle this?

回答1:

The problem isn't with alembic but lies in your usage of MySQL, which can't rollback DDL statements.

So the only (ugly) way to achieve it would be to do manual exception handling and reversing the operations that were successful until that point.

Something like this (written out of my mind, so it's not the most elegant solution and maybe even a little wrong, but I hope you get the gist):

def upgrade():
    try:
        op.create_table('sometable',
            Column('id', INTEGER, primary_key=True),
            Column('name', VARCHAR(150), nullable=False, unique=True))
    except:
        try:
            op.drop_table('sometable')
        except:
            pass
        raise

    try:
        op.add_column('anothertable' Column('id', INTEGER))
    except:
        op.drop_table('sometable')
        try:
            op.drop_column('anothertable', 'id')
        except:
            pass
        raise

    try:
        op.create_table('secondtable')
    except:
        op.drop_table('sometable')
        op.drop_column('anothertable', 'id')
        try:
            op.drop_table('secondtable')
        except:
            pass
        raise

回答2:

If you have in version control the database model you were migrating from, you can temporarily use it to create a migration. (You might have to empty out your versions folder / put everything in a temp directory) Alembic will compare the current state of the database with the model, and give you migration commands for the database to reach that state. In this case, it should give you the instructions for reverting the database to the previous state. You will have to look at the migration commands that were generated to make sure it's exactly what you need, but you won't need to generate them yourself.

After that you can delete the migration, and roll back to the latest db model file. Then you should be back to the point you started

来源：https://stackoverflow.com/questions/17894240/how-to-clean-up-incomplete-alembic-run

标签

python

sqlalchemy

alembic