Kevin Benton 3c8570f3fc Make Neutron resources reference standard attr table

This adds a new 'standardattributes' table and adds a foreign-key
references from ports, subnets, networks, subnetpools, routers,
securitygroups, and floatingips to this table.

This will make it easy to add new things to the schema like
timestamp fields or anything else that applies to multiple types
of Neutron resources. The new fields would just be added to the
'neutronresources' table instead of being duplicated across each
resource's table. Or, if the the relationship is 1-to-many (e.g. tags),
the new association table would be related to the 'standardattribute'
table.

Related-Bug: #1496802
Change-Id: Iaa3ba81a7e9cae09cea153720b29879d8cc9a080

2015-11-23 07:41:26 -08:00

4.2 KiB

Raw Blame History

Neutron Database Layer

This section contains some common information that will be useful for developers that need to do some db changes.

Difference between 'default' and 'server_default' parameters for columns

For columns it is possible to set 'default' or 'server_default'. What is the difference between them and why should they be used?

The explanation is quite simple:

default - the default value that SQLAlchemy will specify in queries for creating instances of a given model;
server_default - the default value for a column that SQLAlchemy will specify in DDL.

Summarizing, 'default' is useless in migrations and only 'server_default' should be used. For synchronizing migrations with models server_default parameter should also be added in model. If default value in database is not needed, 'server_default' should not be used. The declarative approach can be bypassed (i.e. 'default' may be omitted in the model) if default is enforced through business logic.

Database migrations

For details on the neutron-db-manage wrapper and alembic migrations, see Alembic Migrations.

Tests to verify that database migrations and models are in sync

neutron.tests.functional.db.test_migrations

_TestModelsMigrations

The Standard Attribute Table

There are many attributes that we would like to store in the database which are common across many Neutron objects (e.g. tags, timestamps, rbac entries). We have previously been handling this by duplicating the schema to every table via model mixins. This means that a DB migration is required for each object that wants to adopt one of these common attributes. This becomes even more cumbersome when the relationship between the attribute and the object is many-to-one because each object then needs its own table for the attributes (assuming referential integrity is a concern).

To address this issue, the 'standardattribute' table is available. Any model can add support for this table by inheriting the 'HasStandardAttributes' mixin in neutron.db.model_base. This mixin will add a standard_attr_id BigInteger column to the model with a foreign key relationship to the 'standardattribute' table. The model will then be able to access any columns of the 'standardattribute' table and any tables related to it.

The introduction of a new standard attribute only requires one column addition to the 'standardattribute' table for one-to-one relationships or a new table for one-to-many or one-to-zero relationships. Then all of the models using the 'HasStandardAttribute' mixin will automatically gain access to the new attribute.

Any attributes that will apply to every neutron resource (e.g. timestamps) can be added directly to the 'standardattribute' table. For things that will frequently be NULL for most entries (e.g. a column to store an error reason), a new table should be added and joined to in a query to prevent a bunch of NULL entries in the database.

4.2 KiB Raw Blame History