Since you are also using Redshift from Elixir you might be interested that we released RedshiftEcto (Redshift adapter for Ecto) yesterday. It aims to provide a good experience working with a Redshift database from Ecto. Features unsupported by Redshift will raise an error and the adapter supports table and column options specific to Redshift in migrations.

There’s a separate topic for the announcement:

i use postgrex with redshift internally. you need version 0.13.1 or later (not 1.0.0-rc.0 or 1.0.0-rc.1 which predate the 0.13 series) but it should work fine

Hi @demem123,

The system is now in production and works well.

We’ve gone with using Ecto to run stringified/templated SQL against our Redshift cluster. The system is purely internal and only runs trusted queries, so we elected to go this route despite string interpolation in SQL being wildly insecure otherwise.

One thing working in our favor is that the queries we’re running against Redshift using this method are 1. largely administrative, 2. simple, 3. rarely change. For those reasons, we haven’t felt a great need to use Ecto’s query builder or anything on that level of sophistication.

Overall this works well enough for our particular workflow, but I would definitely not use this method if any piece of the system was public facing.

redshift doesn’t support upsert and the redshift client library is actually just postgrex so if you try to use ecto upserts with redshift you’ll get sql errors in your queries

postgrex has pooling built in. see the options for start_link