Do not call setup_db from engine

Calling setup_db is creating the db structure if missing (because it's
calling models.Workbook.metadata.create_all after).

This is not what we want since we want the operator to manage the
creation of the db using mistral-db-manage like any other regular
openstack service.

I suspect that this call was mostly useful for tests.
This is a real issue in production because it create the DB structure
without using the correct alembic migration path (and thus,
alembic_version table is missing).

Change-Id: I51f4128dd023846a4406cd1b414a56d741035483
Signed-off-by: Arnaud M <arnaud.morin@gmail.com>
This commit is contained in:
Arnaud M
2025-09-22 22:32:09 +02:00
parent d2994f8e84
commit 8be59adc38
2 changed files with 0 additions and 6 deletions

View File

@@ -24,7 +24,6 @@ from mistral.api import access_control
from mistral.api.hooks import maintenance
from mistral import config as m_config
from mistral import context as ctx
from mistral.db.v2 import api as db_api_v2
from mistral.rpc import base as rpc
from mistral.services import periodic
@@ -54,8 +53,6 @@ def setup_app(config=None):
app_conf = dict(config.app)
db_api_v2.setup_db()
# TODO(rakhmerov): Why do we run cron triggers in the API layer?
# Should we move it to engine?s
if cfg.CONF.cron_trigger.enabled:

View File

@@ -15,7 +15,6 @@
from oslo_log import log as logging
from mistral import config as cfg
from mistral.db.v2 import api as db_api
from mistral.engine import default_engine
from mistral import exceptions as exc
from mistral.rpc import base as rpc
@@ -60,8 +59,6 @@ class EngineServer(service_base.MistralService):
_validate_config()
db_api.setup_db()
self._scheduler = sched_base.get_system_scheduler()
self._scheduler.start()