How to Persist Orders to DynamoDB From a Python App on AWS EC2

Why put_item returns 500 with a ValidationException

A DynamoDB table is schemaless for its non-key attributes, but the partition key (and sort key, if the table has one) is not optional. Every put_item must include the partition key attribute, named exactly as it was declared when the table was created. Miss it and DynamoDB rejects the write:

ValidationException: One or more parameter values were invalid:
Missing the key order_id in the item

In a typical broken orders service the table was created with order_id as the partition key, but the boto3 write sends id instead. Because the Flask handler wraps the write in a try/except that returns 500, the client only sees a generic error and the table stays empty. Confirm the real table schema before touching code:

aws dynamodb describe-table --table-name Orders \
    --query 'Table.KeySchema'

That prints [{"AttributeName": "order_id", "KeyType": "HASH"}] - the write has to supply order_id, not id.

The fix: match the item key to the table key

The bug lives in the put_item call. The item dict names the partition key id, which the table does not know about:

# BEFORE - write fails with ValidationException, handler returns 500
table.put_item(Item={
    "id": body["id"],
    "customer": body["customer"],
    "amount": body["amount"],
})

Rename the field to the declared partition key. Nothing else changes - the other attributes are free-form:

# AFTER - order_id matches the table's HASH key, write succeeds
table.put_item(Item={
    "order_id": body["id"],
    "customer": body["customer"],
    "amount": body["amount"],
})

Restart the service and re-send the order. A correct write returns 201:

pkill -f order_writer.py; python3 order_writer.py &

curl -sX POST http://localhost:8000/orders \
    -H 'Content-Type: application/json' \
    -d '{"id":"o-1","customer":"alice","amount":42}'
# {"created":"o-1"}  with HTTP 201

Verify the item actually landed in the table instead of trusting the status code:

aws dynamodb scan --table-name Orders --select COUNT --query 'Count'

Keep the writer and the table in sync

This class of bug is silent whenever the exception is swallowed, so the goal is to make a key mismatch impossible rather than to catch it after the fact.

• Single source of truth for key names. Service teams keep the table key names in a shared constants module, or generate them from the same infrastructure-as-code that creates the table (Terraform, CloudFormation) so the writer and the table definition cannot drift apart.
• Validate the item shape before AWS. Type-check the item dict against a TypedDict or a pydantic model so a wrong key name fails locally, not on the wire.
• Do not swallow the write error. If you must catch it, log str(e) at error level - a raw ValidationException names the missing key and turns a mystery 500 into a one-line fix.

The same pattern applies whether the app runs on AWS EC2, in a container, or on Lambda: the partition key attribute must be present in every write, spelled the same way the table was created.