Resource Naming

Stelvio names every AWS resource for you. Knowing the shape helps when you look at the AWS console, write an IAM policy by hand, or wonder why a rename created a new resource.

The shape

<app>-<env>-<component>-<random>

For an app shop, environment prod, and Queue("orders"):

shop-prod-orders-a1b2c3d

The app and environment prefix keeps environments apart in one AWS account. The random 7-character tail is added by Pulumi when the resource is created. It lets a replacement (a setting AWS can't change in place) create the new resource before deleting the old one, with no name collision and no gap where the name doesn't exist.

FIFO queues and topics end in .fifo after the random tail: shop-prod-orders-a1b2c3d.fifo.

Deterministic names

Three resources carry no random tail, because the provider needs the name up front or the name itself is the grouping key:

Long names

AWS limits name length per service: 63 for S3 buckets, 80 for SQS queues, 128 for Cognito user pools. When the prefix plus your component name would overflow, Stelvio keeps the head of the name and replaces the tail with a 7-character hash, so the name stays unique and within the limit. Keep app, environment and component names short and the hash never appears.

Renaming a component

Changing a component's name in stlv_app.py is a new resource to Pulumi, not a rename. The next deploy creates the new one and deletes the old. The app or environment name works the same way, see Renaming. Where that bites:

• Bucket: the old bucket keeps its objects, and the delete fails while it is non-empty. Move or delete the data first.
• UserPool: users live in the pool. The new pool starts empty.
• DynamoTable: the data goes with the table.
• Queue: in-flight messages are lost with the old queue.

Read the stlv diff output before a deploy that shows a delete on any of these.