Running job causes detachedinstanceerror

[bnewman70]

Is there any chance that when running a job using the scheduler within an app context that it can cause a previous session to detach?

i am using version 1.12.3 of flask-apscheduler, I only get the following error when running a job that uses the database. If I run a job that just prints to screen, I do not get this error.

sqlalchemy.orm.exc.DetachedInstanceError: Parent instance <Users at 0x2792f1448b0> is not bound to a Session; lazy load operation of attribute 'profile' cannot proceed.

The call for profile is used by flask-login on @login_required decorator. I have tried many scenarios and this only seems to happen when a job runs. I am using the @scheduler.task decorator to start the job and have my code within this block

with scheduler.app.app_context():

So this works fine
@scheduler.task("interval",id=ups_tracking,seconds=60,max_instances=1,start_date="2000-01-01 12:19:00")
def hello_world():
print("hello world job")

this produces error
@scheduler.task("interval",id=ups_tracking,seconds=60,max_instances=1,start_date="2000-01-01 12:19:00")
def hello_world():
with scheduler.app.app_context():
print("hello world job")

not even doing any db queries
and this only happens when using scheduler.run_job() method on demand.

[christopherpickering]

hey @bnewman70, so if I understand you correctly, the error is only coming up in jobs that have the @login_required decorator on them as well? Can you share code where the decorator is added?

[bnewman70]

the jobs don't have the @login_required decorator, the routes do.
Using Flask-login 0.5.0
When I try to access an object under the current__user that's when the error occurs, but only after I call scheduler.run_job and the job runs within the block
with scheduler.app.app_context():

If I do a run_job in a test job that does not run within a with scheduler.app.app_context(): block, the object under my current_user is correctly pulled from the database with no error.

Heres some code blocks in order of execution:

1. Route for a page that lists all jobs and allows user to perform actions, like pause, resume and run now. Note i added a current_user.profile call as a work around to avoiding the error
   @blueprint.route('/jobs', methods=['GET','POST'])
   @login_required
   def jobs():
   action = request.args.get('action')
   jobs = scheduler.get_jobs()
   print('got jobs',len(jobs))
   if action == 'pause':
   print('pausing job')
   scheduler.pause_job(request.args.get('jid'))
   flash('Job paused',category='success')
   elif action == 'resume':
   scheduler.resume_job(request.args.get('jid'))
   flash('Job Resumed',category='success')
   elif action == 'run_now':
   # avoid detached profile only in run now, pull profile before job runs
   print(current_user.profile)
   scheduler.run_job(request.args.get('jid'))
   flash('Job Started',category='success')

2. example job, triggering the detached instance error. If I remove with scheduer.app.app_context(), the error does not occur
   @scheduler.task("interval",id=hello_world,seconds=60,max_instances=1,start_date="2000-01-01 12:19:00")
   def hello_world():
   with scheduler.app.app_context():
   print("hello world job")

3. In my template I show the user profile image from the current_user.profile in the header of each page.. this triggers the error. I an guessing I only see it on run_now because this page immediately reloads after run. Other pages may not be refreshed during a job run.
   {% if current_user.profile %}
   {% for user in current_user.profile %}
   
   {% endfor %}
   {% endif %}
   Thanks

[christopherpickering]

It is difficult to follow, maybe if you have code on github somewhere it would be easier.

It sounds as though you have a db session, and after you pass it into a job, the session is no longer accessible outside the job.

Are you using the extensions pattern (similar to this example) to create you db?

You could have something like

from flask_sqlalchemy import SQLAlchemy
from flask_sqlalchemy_caching import CachingQuery

db = SQLAlchemy(query_class=CachingQuery)

in your extensions.py and then import it where you need db connections, and use it inside the app_context().

In your init.py you would still need to also import db and initialize it with db.init_app(app)