Registry migrate process lock in PostgreSQL

[Mathieu GUILLAUME]

I started the registry on an empty db (sqlalchemy connector: postgresql+psycopg2). It gets stuck on migrate phase 2->3 (same thing if I restart):

2011-07-01 17:09:24 INFO [migrate.versioning.api] 2 -> 3...
2011-07-01 17:09:24 INFO [sqlalchemy.engine.base.Engine.0x...9b10] SELECT images.id, images.name, images.type, images.size, images.status, images.is_public, images.location, images.created_at, images.updated_at, images.deleted_at, images.deleted
FROM images LEFT OUTER JOIN image_properties ON images.id = image_properties.image_id AND image_properties.key = %(key_1)s
WHERE image_properties.image_id IS NULL AND images.type IS NOT NULL
2011-07-01 17:09:24 INFO [sqlalchemy.engine.base.Engine.0x...9b10] {'key_1': 'type'}
2011-07-01 17:09:24 DEBUG [sqlalchemy.engine.base.Engine.0x...9b10] Col ('id', 'name', 'type', 'size', 'status', 'is_public', 'location', 'created_at', 'updated_at', 'deleted_at', 'deleted')
2011-07-01 17:09:24 INFO [sqlalchemy.engine.base.Engine.0x...9b10]
ALTER TABLE images ADD disk_format VARCHAR(20)
2011-07-01 17:09:24 INFO [sqlalchemy.engine.base.Engine.0x...9b10] {}

If I look at the locks in the database, I can see that the "alter" is waiting on an AccessExclusiveLock for 'images'. There is an existing AccessShareLock on the table, I guess that's why the exclusive lock is not granted.

[Jay Pipes (jaypipes)]

Hi Matthieu,

Sorry for the long delay in getting to this bug...

So, it seems to me that this is a bug in the upstream project (sqlalchemy-migrate). Would it be possible to submit a bug report for that project?

http://code.google.com/p/sqlalchemy-migrate/issues/list

Thanks!
-jay