Query in Athena

Data Pump catalogs the result in the Glue Data Catalog. Any tool that reads the catalog — Athena, Redshift Spectrum, EMR, Spark — sees the tables.

How the tables appear

The name comes from the database and the table, with the aliases you configured:

In the database	In the data lake	In Athena
pagila / public.payment	pagila.payment	pagila_payment
with aliases vendas / pagamentos	vendas.pagamentos	vendas_pagamentos

The public schema is removed; other schemas are preserved.

Query

SELECT customer_id, sum(CAST(amount AS DECIMAL(10,2))) AS total
  FROM datalake.vendas_pagamentos
 WHERE ano = 2024 AND mes = 3
 GROUP BY customer_id
 ORDER BY total DESC
 LIMIT 20;

The WHERE on the partition columns is what makes the difference in cost: without it, Athena reads the whole table.

Check how much was read

Athena shows the volume scanned for each query. If filtering by partition does not reduce that number, the partition is not being used — check that the column name in the WHERE is the same one that appears under Datalake.

Types

The RDS export delivers several columns as text, including dates and decimal numbers. That comes from the format, not from Data Pump.

-- data
CAST(payment_date AS TIMESTAMP)

-- decimal
CAST(amount AS DECIMAL(10,2))

If a CAST is used every time, it is worth creating an extra column in the partition configuration: it is computed once, at ingestion, and arrives ready in Athena.

See what exists

The console’s Datalake screen lists the cataloged tables with columns, types, row counts and the source of each one. It is faster than querying the catalog through Athena when you only want to know what is there.