Files were the warm-up. Most pipelines end in a database, and this world gives you a real one: a managed PostgreSQL, free, reachable over the internet with SSL — the same shape as the database you'll meet at work. You'll save the connection once, then load the course's three CSV files into it with Odara itself.
Goal
A saved connection called academy_pg, and three tables in your database — customers (500 rows),
products (8) and orders (3,000) — loaded by one pipeline.
You'll need
- A free PostgreSQL database. Two providers that work well, both without a credit card at the time of writing: Aiven (free PostgreSQL plan) and Neon. Plans change — check the provider's current terms. Anything that speaks PostgreSQL 15 or newer works.
customers.csv,products.csvandorders.csvfrom World 3 (level 3.2's download).- About 25 minutes.
From your provider you need six things, usually shown together as a connection string or Service URI: host, port, database, user, password, and the fact that it requires SSL. For example:
postgres://avnadmin:[email protected]:12345/defaultdb?sslmode=require
└ user ┘ └ pass ┘ └──────────────── host ──────────────┘ └port┘ └ database ┘
A new database can take a few minutes before its address exists — if Odara says the host can't be found, wait until the provider shows the service as running.
Steps
1. Save the connection
Open Resources in the left panel. Under DATA STORES, hover PostgreSQL and click +. The New Metadata form opens with Database Connection and PostgreSQL already selected:
| Field | Value |
|---|---|
| Name | academy_pg |
| Host, Port, Database, Username, Password | from your connection string |
| Schema | leave empty (public) |
| SSL Mode | require |
Click Test Connection. You should see Connection successful (…ms) - PostgreSQL 17… (or whichever
version your provider runs). Then Create.

SSL Mode defaults to prefer, which will use SSL if the server offers it. Managed databases
require it, so say so: with require, a connection that somehow isn't encrypted fails instead of
silently going through.
2. Load three files, one pipeline
Create a pipeline w4-1-load with three CSV Source nodes — customers, products, orders —
and three PostgreSQL Write nodes (DELIVER ▸ Databases), one after each.

Click each PostgreSQL Write and pick academy_pg in Use Connection — the connection details fill
themselves in. Then:
| Field | Value |
|---|---|
| Table Name | customers / products / orders |
| Table Operation (DDL) | Drop & Create |
| Data Operation (DML) | Insert |

The screenshots write into a schema called academy because the author's database is shared. In your
own database, public is fine and your queries don't need a prefix.
Drop & Create deletes the table if it exists and creates it again from the incoming columns, so every run starts clean. Insert then writes the rows.
3. Execute
Execute. The three pairs run side by side; the whole load takes a couple of seconds.
Expected result
The log shows
load_customers: 500 rows,load_products: 8 rows,load_orders: 3000 rows.The tables have the types Odara inferred from the files:
Column PostgreSQL type order_id,customer_id,quantitybigintorder_date,signup_datedatemonthly_spend,list_pricedouble precisionactivebooleaneverything else textRun it again: same counts. Drop & Create makes the load repeatable.
Common pitfalls
failed to lookup address information: Name or service not known.
The host doesn't exist (yet). Check for a typo, or wait for a brand-new database to finish starting.
no pg_hba.conf entry for host "…", user "…", database "…", no encryption.
SSL is off — SSL Mode is disable. Managed databases refuse unencrypted connections. Set it to
require.
password authentication failed for user "…".
Wrong password, or the right password for another user. Copy it again from the provider.
pool timed out while waiting for an open connection.
Nothing answered on that port — very often the default 5432 left in place when the provider uses a
different port. Check the port in the connection string.
Drop & Create on someone else's table. It really drops. Point a course pipeline only at a database, or a schema, that's yours to experiment in.
Sharing a pipeline that uses a connection.
Exported pipeline JSON never contains passwords — they're replaced by ****MASKED****. The node
keeps a reference to your saved connection, so importing the file into a project that has that
connection just works: the password comes from the connection when the pipeline runs. Imported
somewhere without it, the import message names the nodes that need a connection chosen (or the
password typed again), and running before that fails with has a hidden password.
What you built
A database with the course data in it, loaded by a pipeline you can rerun at any time. The next levels read from it, write to it in the other modes, and apply changes to it.
Check yourself
One question, drawn from a larger bank. Get it right to pass.
Why set SSL Mode to `require` instead of the default `prefer`?