Flask/Heroku ClearDB not updating

Question

I've deployed my Python app on Heroku and have connected it to a clearDB database. This works perfectly when run locally, however when running on Heroku my INSERT statements don't seem to be committing properly (sometimes they do, sometimes they don't). It's running through gunicorn.

This is my MySQL setup:

APP_ROOT = os.path.join(os.path.dirname(__file__), '..')
dotenv_path = os.path.join(APP_ROOT, '.env')
load_dotenv(dotenv_path)

db_user = os.getenv('DB_USER2')
db_pass = os.getenv('DB_PASSWORD2')
db_database = os.getenv('DB_DATABASE2')
db_host = os.getenv('DB_HOST2')

app.config['MYSQL_DATABASE_USER'] = db_user
app.config['MYSQL_DATABASE_PASSWORD'] = db_pass
app.config['MYSQL_DATABASE_DB'] = db_database
app.config['MYSQL_DATABASE_HOST'] = db_host
mysql.init_app(app)

These variables are coming from a .env file. But they are also added as environmental variables on Heroku. So connection to the server is same both locally and on Heroku.

It's completely random whether it works or not. Fetch statements work perfectly fine, but inserting or updating seems to cause me a lot of issues. Again, this works perfectly with no issues locally, but doesn't on Heroku. (When run locally, it seems to take a bit of time to save the data, which is fine. But it doesn't take much time at all on Heroku so I'm assuming it's somehow skipping the commit).

There are no errors in the logs either. It just seems to bypass the statement sometimes and not commit. Doing a hard refresh sometimes brings up the data, but then it disappears again. Most of the times the data isn't saved at all.

I have one cursor open throughout the app and I'm using conn.ping(True) before every SQL statement, and conn.commit() after fetching the data.

Any help would be appreciated, thank you!

I updated the post. They're just variables coming from a .env file. But they're set up on Heroku as environmental variables too. — dibsonme, Dec 07 '17 at 23:15

score 2 · Answer 1 · edited Jun 05 '19 at 10:55

Heroku uses two masters. so sometimes your credential are available and sometimes not, in your procfile, you must have

web: gunicorn --workers=1 app:app

workers=1 means your data will be on one master, app=app is just my entry point, yours can be what you have.

Below is from Heroku Dual Masters, One Database Connection URL The ClearDB Heroku Add-On provisions your database on two "multi-master" database servers in two different Amazon EC2 regions to ensure absolute data availability. In our 1.0 add-on, this meant that you received two database connections. Now, with our custom SQL high availability routing technology, you get a single connection that automatically fails over to the secondary endpoint in the event of a failure, and will switch back to the primary master once service has been restored and the databases have re-synchronized.

The primary master node is located right next to Heroku, in Amazon's EC2 "US-East" region. The secondary master node is located in Amazon's EC2 "US-West" region for high availability purposes.

Flask/Heroku ClearDB not updating

1 Answers1