Fork me on GitHub

Resource properties

Each resource object has two kinds of properties: inputs and outputs.

At creation time, you supply input properties to control the desired state of the resource:

const bucket = new aws.s3.Bucket("photos", { versioning: { enable: true } });
let bucket = new aws.s3.Bucket("photos", { versioning: { enable: true } });
bucket = s3.Bucket('my-bucket', versioning={ 'enable': True })
_, err := s3.NewBucket(ctx, "my-bucket", &s3.BucketArgs{
    Versioning: map[string]interface{}{
        "enable": true,

Inputs properties are key to how Pulumi works; the diffing of them instructs Pulumi how to do its deployments. And they are what the resource provider uses to perform actual creations and updates.

The resulting resource object offers output properties that communicate its final state:

bucket.bucketDomainName.apply(function (name) {
bucket.bucketDomainName.apply((name: string) => {
bucket.BucketDomainName.Apply(func (name string) error {
    return nil

A resource often has different output properties than input properties, because cloud providers may compute outputs. For instance, the URL for an API, the IP address of a load balancer, etc, are typically auto-assigned.

A resource’s output properties are promise-like in nature, because they aren’t known until creation finishes. Hence the use of the “apply”-style functions above. Outputs will, in fact, not even resolve to a value during previews.

Output properties also carry dependency information, so that if an output from one resource is supplied as input to another, Pulumi can track dependencies accurately for purposes of parallelism, safe deletion order, and visualizations.