oslo.db/oslo_db/sqlalchemy/provision.py
Jeremy Stanley 9b552046f5 Switch from MySQL-python to PyMySQL
As discussed in the Liberty Design Summit "Moving apps to Python 3"
cross-project workshop, the way forward in the near future is to
switch to the pure-python PyMySQL library as a default.

Added a special test environment to keep MySQL-python support.
Documentation modified.

https://etherpad.openstack.org/p/liberty-cross-project-python3

Change-Id: I12b32dc097a121bd43991bc38dd4d289b65e86c1
2015-06-18 15:42:23 +03:00

578 lines
19 KiB
Python

# Copyright 2013 Mirantis.inc
# All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
"""Provision test environment for specific DB backends"""
import abc
import logging
import os
import random
import re
import string
import six
from six import moves
import sqlalchemy
from sqlalchemy.engine import url as sa_url
from sqlalchemy import schema
import testresources
from oslo_db._i18n import _LI
from oslo_db import exception
from oslo_db.sqlalchemy.compat import utils as compat_utils
from oslo_db.sqlalchemy import session
from oslo_db.sqlalchemy import utils
LOG = logging.getLogger(__name__)
class ProvisionedDatabase(object):
pass
class BackendResource(testresources.TestResourceManager):
def __init__(self, database_type):
super(BackendResource, self).__init__()
self.database_type = database_type
self.backend = Backend.backend_for_database_type(self.database_type)
def make(self, dependency_resources):
return self.backend
def isDirty(self):
return False
class DatabaseResource(testresources.TestResourceManager):
def __init__(self, database_type):
super(DatabaseResource, self).__init__()
self.database_type = database_type
self.resources = [
('backend', BackendResource(database_type))
]
def make(self, dependency_resources):
dependency_resources['db_token'] = db_token = _random_ident()
backend = dependency_resources['backend']
LOG.info(
"CREATE BACKEND %s TOKEN %s", backend.engine.url, db_token)
backend.create_named_database(db_token, conditional=True)
dependency_resources['engine'] = \
backend.provisioned_engine(db_token)
return ProvisionedDatabase()
def clean(self, resource):
resource.engine.dispose()
LOG.info(
"DROP BACKEND %s TOKEN %s",
resource.backend.engine, resource.db_token)
resource.backend.drop_named_database(resource.db_token)
def isDirty(self):
return False
class TransactionResource(testresources.TestResourceManager):
def __init__(self, database_resource, schema_resource):
super(TransactionResource, self).__init__()
self.resources = [
('database', database_resource),
('schema', schema_resource)
]
def clean(self, resource):
resource._dispose()
def make(self, dependency_resources):
conn = dependency_resources['database'].engine.connect()
return utils.NonCommittingEngine(conn)
def isDirty(self):
return True
class Schema(object):
pass
class SchemaResource(testresources.TestResourceManager):
def __init__(self, database_resource, generate_schema, teardown=False):
super(SchemaResource, self).__init__()
self.generate_schema = generate_schema
self.teardown = teardown
self.resources = [
('database', database_resource)
]
def clean(self, resource):
LOG.info(
"DROP ALL OBJECTS, BACKEND %s",
resource.database.engine.url)
resource.database.backend.drop_all_objects(
resource.database.engine)
def make(self, dependency_resources):
if self.generate_schema:
self.generate_schema(dependency_resources['database'].engine)
return Schema()
def isDirty(self):
if self.teardown:
return True
else:
return False
class Backend(object):
"""Represent a particular database backend that may be provisionable.
The ``Backend`` object maintains a database type (e.g. database without
specific driver type, such as "sqlite", "postgresql", etc.),
a target URL, a base ``Engine`` for that URL object that can be used
to provision databases and a ``BackendImpl`` which knows how to perform
operations against this type of ``Engine``.
"""
backends_by_database_type = {}
def __init__(self, database_type, url):
self.database_type = database_type
self.url = url
self.verified = False
self.engine = None
self.impl = BackendImpl.impl(database_type)
self.current_dbs = set()
Backend.backends_by_database_type[database_type] = self
@classmethod
def backend_for_database_type(cls, database_type):
"""Return the ``Backend`` for the given database type.
"""
try:
backend = cls.backends_by_database_type[database_type]
except KeyError:
raise exception.BackendNotAvailable(database_type)
else:
return backend._verify()
@classmethod
def all_viable_backends(cls):
"""Return an iterator of all ``Backend`` objects that are present
and provisionable.
"""
for backend in cls.backends_by_database_type.values():
try:
yield backend._verify()
except exception.BackendNotAvailable:
pass
def _verify(self):
"""Verify that this ``Backend`` is available and provisionable.
:return: this ``Backend``
:raises: ``BackendNotAvailable`` if the backend is not available.
"""
if not self.verified:
try:
eng = self._ensure_backend_available(self.url)
except exception.BackendNotAvailable:
raise
else:
self.engine = eng
finally:
self.verified = True
if self.engine is None:
raise exception.BackendNotAvailable(self.database_type)
return self
@classmethod
def _ensure_backend_available(cls, url):
url = sa_url.make_url(str(url))
try:
eng = sqlalchemy.create_engine(url)
except ImportError as i_e:
# SQLAlchemy performs an "import" of the DBAPI module
# within create_engine(). So if ibm_db_sa, cx_oracle etc.
# isn't installed, we get an ImportError here.
LOG.info(
_LI("The %(dbapi)s backend is unavailable: %(err)s"),
dict(dbapi=url.drivername, err=i_e))
raise exception.BackendNotAvailable("No DBAPI installed")
else:
try:
conn = eng.connect()
except sqlalchemy.exc.DBAPIError as d_e:
# upon connect, SQLAlchemy calls dbapi.connect(). This
# usually raises OperationalError and should always at
# least raise a SQLAlchemy-wrapped DBAPI Error.
LOG.info(
_LI("The %(dbapi)s backend is unavailable: %(err)s"),
dict(dbapi=url.drivername, err=d_e)
)
raise exception.BackendNotAvailable("Could not connect")
else:
conn.close()
return eng
def create_named_database(self, ident, conditional=False):
"""Create a database with the given name."""
if not conditional or ident not in self.current_dbs:
self.current_dbs.add(ident)
self.impl.create_named_database(
self.engine, ident, conditional=conditional)
def drop_named_database(self, ident, conditional=False):
"""Drop a database with the given name."""
self.impl.drop_named_database(
self.engine, ident,
conditional=conditional)
self.current_dbs.discard(ident)
def drop_all_objects(self, engine):
"""Drop all database objects.
Drops all database objects remaining on the default schema of the
given engine.
"""
self.impl.drop_all_objects(engine)
def database_exists(self, ident):
"""Return True if a database of the given name exists."""
return self.impl.database_exists(self.engine, ident)
def provisioned_engine(self, ident):
"""Given the URL of a particular database backend and the string
name of a particular 'database' within that backend, return
an Engine instance whose connections will refer directly to the
named database.
For hostname-based URLs, this typically involves switching just the
'database' portion of the URL with the given name and creating
an engine.
For URLs that instead deal with DSNs, the rules may be more custom;
for example, the engine may need to connect to the root URL and
then emit a command to switch to the named database.
"""
return self.impl.provisioned_engine(self.url, ident)
@classmethod
def _setup(cls):
"""Initial startup feature will scan the environment for configured
URLs and place them into the list of URLs we will use for provisioning.
This searches through OS_TEST_DBAPI_ADMIN_CONNECTION for URLs. If
not present, we set up URLs based on the "opportunstic" convention,
e.g. username+password = "openstack_citest".
The provisioning system will then use or discard these URLs as they
are requested, based on whether or not the target database is actually
found to be available.
"""
configured_urls = os.getenv('OS_TEST_DBAPI_ADMIN_CONNECTION', None)
if configured_urls:
configured_urls = configured_urls.split(";")
else:
configured_urls = [
impl.create_opportunistic_driver_url()
for impl in BackendImpl.all_impls()
]
for url_str in configured_urls:
url = sa_url.make_url(url_str)
m = re.match(r'([^+]+?)(?:\+(.+))?$', url.drivername)
database_type, drivertype = m.group(1, 2)
Backend(database_type, url)
@six.add_metaclass(abc.ABCMeta)
class BackendImpl(object):
"""Provide database-specific implementations of key provisioning
functions.
``BackendImpl`` is owned by a ``Backend`` instance which delegates
to it for all database-specific features.
"""
default_engine_kwargs = {}
supports_drop_fk = True
@classmethod
def all_impls(cls):
"""Return an iterator of all possible BackendImpl objects.
These are BackendImpls that are implemented, but not
necessarily provisionable.
"""
for database_type in cls.impl.reg:
if database_type == '*':
continue
yield BackendImpl.impl(database_type)
@utils.dispatch_for_dialect("*")
def impl(drivername):
"""Return a ``BackendImpl`` instance corresponding to the
given driver name.
This is a dispatched method which will refer to the constructor
of implementing subclasses.
"""
raise NotImplementedError(
"No provision impl available for driver: %s" % drivername)
def __init__(self, drivername):
self.drivername = drivername
@abc.abstractmethod
def create_opportunistic_driver_url(self):
"""Produce a string url known as the 'opportunistic' URL.
This URL is one that corresponds to an established Openstack
convention for a pre-established database login, which, when
detected as available in the local environment, is automatically
used as a test platform for a specific type of driver.
"""
@abc.abstractmethod
def create_named_database(self, engine, ident, conditional=False):
"""Create a database with the given name."""
@abc.abstractmethod
def drop_named_database(self, engine, ident, conditional=False):
"""Drop a database with the given name."""
def drop_all_objects(self, engine):
"""Drop all database objects.
Drops all database objects remaining on the default schema of the
given engine.
Per-db implementations will also need to drop items specific to those
systems, such as sequences, custom types (e.g. pg ENUM), etc.
"""
with engine.begin() as conn:
inspector = sqlalchemy.inspect(engine)
metadata = schema.MetaData()
tbs = []
all_fks = []
for table_name in inspector.get_table_names():
fks = []
for fk in inspector.get_foreign_keys(table_name):
# note that SQLite reflection does not have names
# for foreign keys until SQLAlchemy 1.0
if not fk['name']:
continue
fks.append(
schema.ForeignKeyConstraint((), (), name=fk['name'])
)
table = schema.Table(table_name, metadata, *fks)
tbs.append(table)
all_fks.extend(fks)
if self.supports_drop_fk:
for fkc in all_fks:
conn.execute(schema.DropConstraint(fkc))
for table in tbs:
conn.execute(schema.DropTable(table))
self.drop_additional_objects(conn)
def drop_additional_objects(self, conn):
pass
def provisioned_engine(self, base_url, ident):
"""Return a provisioned engine.
Given the URL of a particular database backend and the string
name of a particular 'database' within that backend, return
an Engine instance whose connections will refer directly to the
named database.
For hostname-based URLs, this typically involves switching just the
'database' portion of the URL with the given name and creating
an engine.
For URLs that instead deal with DSNs, the rules may be more custom;
for example, the engine may need to connect to the root URL and
then emit a command to switch to the named database.
"""
url = sa_url.make_url(str(base_url))
url.database = ident
return session.create_engine(
url,
logging_name="%s@%s" % (self.drivername, ident),
**self.default_engine_kwargs
)
@BackendImpl.impl.dispatch_for("mysql")
class MySQLBackendImpl(BackendImpl):
default_engine_kwargs = {'mysql_sql_mode': 'TRADITIONAL'}
def create_opportunistic_driver_url(self):
return "mysql+pymysql://openstack_citest:openstack_citest@localhost/"
def create_named_database(self, engine, ident, conditional=False):
with engine.connect() as conn:
if not conditional or not self.database_exists(conn, ident):
conn.execute("CREATE DATABASE %s" % ident)
def drop_named_database(self, engine, ident, conditional=False):
with engine.connect() as conn:
if not conditional or self.database_exists(conn, ident):
conn.execute("DROP DATABASE %s" % ident)
def database_exists(self, engine, ident):
return bool(engine.scalar("SHOW DATABASES LIKE '%s'" % ident))
@BackendImpl.impl.dispatch_for("sqlite")
class SQLiteBackendImpl(BackendImpl):
supports_drop_fk = False
def create_opportunistic_driver_url(self):
return "sqlite://"
def create_named_database(self, engine, ident, conditional=False):
url = self._provisioned_database_url(engine.url, ident)
filename = url.database
if filename and (not conditional or not os.access(filename, os.F_OK)):
eng = sqlalchemy.create_engine(url)
eng.connect().close()
def provisioned_engine(self, base_url, ident):
return session.create_engine(
self._provisioned_database_url(base_url, ident))
def drop_named_database(self, engine, ident, conditional=False):
url = self._provisioned_database_url(engine.url, ident)
filename = url.database
if filename and (not conditional or os.access(filename, os.F_OK)):
os.remove(filename)
def database_exists(self, engine, ident):
url = self._provisioned_database_url(engine.url, ident)
filename = url.database
return not filename or os.access(filename, os.F_OK)
def _provisioned_database_url(self, base_url, ident):
if base_url.database:
return sa_url.make_url("sqlite:////tmp/%s.db" % ident)
else:
return base_url
@BackendImpl.impl.dispatch_for("postgresql")
class PostgresqlBackendImpl(BackendImpl):
def create_opportunistic_driver_url(self):
return "postgresql://openstack_citest:openstack_citest"\
"@localhost/postgres"
def create_named_database(self, engine, ident, conditional=False):
with engine.connect().execution_options(
isolation_level="AUTOCOMMIT") as conn:
if not conditional or not self.database_exists(conn, ident):
conn.execute("CREATE DATABASE %s" % ident)
def drop_named_database(self, engine, ident, conditional=False):
with engine.connect().execution_options(
isolation_level="AUTOCOMMIT") as conn:
self._close_out_database_users(conn, ident)
if conditional:
conn.execute("DROP DATABASE IF EXISTS %s" % ident)
else:
conn.execute("DROP DATABASE %s" % ident)
def drop_additional_objects(self, conn):
enums = compat_utils.get_postgresql_enums(conn)
for e in enums:
conn.execute("DROP TYPE %s" % e)
def database_exists(self, engine, ident):
return bool(
engine.scalar(
sqlalchemy.text(
"select datname from pg_database "
"where datname=:name"), name=ident)
)
def _close_out_database_users(self, conn, ident):
"""Attempt to guarantee a database can be dropped.
Optional feature which guarantees no connections with our
username are attached to the DB we're going to drop.
This method has caveats; for one, the 'pid' column was named
'procpid' prior to Postgresql 9.2. But more critically,
prior to 9.2 this operation required superuser permissions,
even if the connections we're closing are under the same username
as us. In more recent versions this restriction has been
lifted for same-user connections.
"""
if conn.dialect.server_version_info >= (9, 2):
conn.execute(
sqlalchemy.text(
"select pg_terminate_backend(pid) "
"from pg_stat_activity "
"where usename=current_user and "
"pid != pg_backend_pid() "
"and datname=:dname"
), dname=ident)
def _random_ident():
return ''.join(
random.choice(string.ascii_lowercase)
for i in moves.range(10))
Backend._setup()