Database guides

Aurora DSQL

Updated 2026-08-01 · 3 min read

AWS’s serverless distributed SQL, PostgreSQL-compatible on the wire. AddisDB connects with the Postgres driver; the catalog it exposes is deliberately small.

Aurora DSQL is a serverless, distributed SQL database from AWS that speaks the PostgreSQL wire protocol. AddisDB connects with its Postgres driver — but DSQL is a fresh implementation rather than a fork, so it is compatible on the wire and with the SQL, not with everything a Postgres server usually exposes underneath.

Who it is for

DSQL suits applications that want Postgres SQL with no instances to size, patch or fail over, and active-active reads and writes across regions. There is no primary to promote, because there is no primary.

The trade is real constraints: no extensions, a restricted set of DDL, optimistic concurrency that surfaces as retryable conflicts at commit rather than as lock waits, and no server you can log into. If your application already handles retries, that is a small tax; if it assumes a session it can hold state in, it is not.

Set up the cluster

  1. Create a cluster in the Aurora DSQL console and copy its endpoint hostname.
  2. Generate an authentication token with the AWS CLI or SDK — DSQL authenticates with a short-lived IAM token, not a stored password.
  3. Note that the token expires (15 minutes by default), so you will regenerate it rather than save it once.
  4. The IAM identity generating the token needs the DSQL connect permission for the cluster, and the admin variant if you connect as admin.
# Mint a token to paste into the password field
aws dsql generate-db-connect-admin-auth-token \
  --hostname <cluster-endpoint> --region us-east-1

Connect from AddisDB

  1. New Connection → Aurora DSQL. Port prefills to 5432, database to postgres, username to admin.
  2. Paste the cluster endpoint as the host.
  3. Paste the generated IAM auth token into the password field.
  4. Leave SSL mode at require — DSQL is TLS-only.
  5. Test, then Save.

What AddisDB gives you

  • Schema browsing, the relationship diagram, in-grid editing, mock data, and AI querying.
  • Introspection runs in lenient mode, so the reduced catalog does not fail a schema refresh.
  • The Chart view, ⌘K search, and destructive-statement detection before anything runs.

Writing SQL DSQL accepts

The restrictions are worth reading before you plan a migration onto it rather than after: no extensions at all, primary keys that cannot be altered after creation, no foreign keys, and a set of types narrower than stock Postgres. Statements outside that envelope come back as the server’s own error, not as anything AddisDB blocked.