Failed SQLA operations don't roll back in dashboard, block threads

[joefutrelle]

A teardown handler can be used to roll back the transaction and release the connection. It still will not reattempt the transaction, and the web request will still fail, but that is better than leaving the connection blocked.

http://stackoverflow.com/questions/33284334/how-to-make-flask-sqlalchemy-automatically-rollback-the-session-if-an-exception

http://stackoverflow.com/questions/21078696/why-is-my-scoped-session-raising-an-attributeerror-session-object-has-no-attr

[joefutrelle]

Probably need a teardown on the workflow web service as well.

[joefutrelle]

Weirdly, adding a teardown on the workflow webapi blueprint causes it to not work at all, so I may not be doing that right.

[joefutrelle]

This has to do with session management. Normally in this situation people use Flask-SQLAlchemy to automatically set the scoped_session sesssion scope to per-request; I was doing it globally.

If you're not using Flask-SQLAlchemy to do this you have to do it yourself

http://stackcode.xyz/sc?id=is37844704