How to change database to Postges in JupyterHub? - postgresql

Question

I am trying to run jupyterhub with my config and I would like to change database form SQLite, that is created by default to PostgreSQL that alredy exists and has some tables (jupyterhub and other app would work concurrently and share database). On website only thing I see is:
We recommend using PostgreSQL for production if you are unsure ...
But no word how to change this database. Have you done this before and can describe it? Like do I ahve to create some tables on my own or do I just pass a link and jupyterhub willl do the rest?

Answer 1

You need create an empty database in postgres and then set the db_url property in the jupyterhub config. For example, for a postgres database on the local machine:
Connect to your postgres instance with a user that has the 'Create DB' attribute and run:
CREATE DATABASE jupyterhub1;
In your jupyterhub_config.py file set this property:
c.JupyterHub.db_url = 'postgresql://username:password#localhost:5432/jupyterhub1'
When you start jupyterhub it will create the required tables automatically. Also note that you don't have to hard code the credentials into the db_url property, you could access them from environment variables using os.environ["VAR_NAME"]
Thanks

Answer 2

Well, I will share what have worked for me. I hope this instruction help other people that face the same difficulty also.
Make a full backup of your database, just in case things go bad. Source
If you read the source above you got that you need to add some fields to your config.yaml.
you'll have to add this part to your config.yaml:
hub:
db:
upgrade: true
Keep following the source mentioned above and you will understand, because there are a little few details to do.
It is not that all, yet! You need to create your database(using Mysql or Postgres). After you have done your database, now you should have username, password, database_name, host and port.
Now, have a look on this info. This another page highlights one interesting point. JupyterHub doc says that using Postgres is easier than Mysql. You 'll have to add hub.db.type and hub.db.url. Have a look on doc to understand the string connection.
postgresql+psycopg2://<db-username>:<db-password>#<db-hostname>:<db-port>/<db-name>
Pay attention that in this example I just put the username and put nothing as password within string. Once it is possible to add hub.db.password I used this option. I also declared the size of database as 20Gi, in my example case.
hub:
db:
upgrade: true
type: postgres
url: postgresql+psycopg2://postgres:#db_jupyterhub_xxxxx.amazonaws.com:5432/db_jupyterhub
password: ~
pvc:
accessModes:
- ReadWriteMany
storage: 20Gi
At this point you realized that I did not put the password. In this case when I will deploy JupyterHub(or make a helm upgrade ...) I need to pass the --set parameter. So for this example I used this way to upgrade the helm chart that already existed.
helm upgrade -f config.yaml jupyterhub . \
--set hub.db.password=TYPE-PASSWORD-OF-DATABASE
With these steps the deploy should work. I used in this way and it worked fine. I have also another point to highlight, cookieSecret. In the docs they mentioned the needs of recreate it in case of pods restarting. Please have a look in this topic about deafult behavior of jupyterhub_cookie_secret and in this link explaining about cookie generation and uses.
I hope these steps help some of you.