Open HeliosH opened 6 years ago
This issue is affecting my project. Anyone have a work-around?
+1
bac and I figured out a way to get some of this working cleanly for the 0.17.x release but you have to add a preprocessor step, and we had to modify the session code a bit. Hopefully for the 1.0 branch we can work on fixing some of this as well. The root of the problem is that a single "session" reference can go stale, and even if your app is using pooling or cleaning up the session it will have failed that 1 request (randomly or based on MySQL connection kills). That means the .session on all your APIs will go bad without some way of resetting it.
https://github.com/UberMeatShield/flask-restless/tree/0.17.1 . I am trying to figure out how to contribute this back but the 0.17 tag is super stale, and I cannot currently figure out all the versions of python / requirements that will get the tests to pass. Working on that, but hopefully having a branch that can work helps somebody if I cannot get a MR for 0.17 in a sensible state.
# This kind of code should be added as a pre-processor, or added around your creation of a session
def set_manager(api_manager):
global API_MANAGER # Reference to your APIManager instance
API_MANAGER = api_manager
def get_manager():
return API_MANAGER
def check_session(**kwargs):
""" Call this in a preprocessor function after you check auth to make sure you session is _actually_ working """
manager = get_manager()
if manager:
try:
session = Session()
session.rollback()
session.execute("SELECT 1")
except Exception as error:
current_app.logger.error("check session: session error: %s", error)
try:
session.remove()
except Exception as remove_err:
current_app.logger.error("Failed to remove session")
current_app.logger.exception(remove_err)
manager.reset_session(Session())
the issue still open.
use these way to create manager manager_restless = APIManager(app, session=data_dict_session)
use blueprint to create api blueprint_table = manager_restless.create_api_blueprint(***) app.register_blueprint(blueprint_table)
add postprocessors to close session
def update_result_format(result, search_params=None, **kw): session.remove()
but it still throw the Exception:
unexpected internal error occurs: (sqlalchemy.exc.InvalidRequestError) Can't reconnect until invalid transaction is rolled back [SQL: u'SELECT count(*) AS count_1 \nFROM dw_table \nWHERE dw_table.id != %s'] [parameters: [{}]]
version: 0.17.0