Data cools as it ages: this week's logs are read daily, last year's almost never. A lifecycle rule moves objects to colder classes as they age, and deletes them at the end.

The rule

{
  "Rules": [{
    "ID": "logs-age",
    "Status": "Enabled",
    "Filter": { "Prefix": "logs/" },
    "Transitions": [
      { "Days": 30, "StorageClass": "STANDARD_IA" },
      { "Days": 90, "StorageClass": "GLACIER" }
    ],
    "Expiration": { "Days": 365 },
    "NoncurrentVersionExpiration": { "NoncurrentDays": 30 }
  }]
}

• Filter limits the rule to a prefix; { "Prefix": "" } is the whole bucket.
• Transitions move the object when it reaches each age.
• Expiration deletes the current version at that age.
• NoncurrentVersionExpiration deletes old versions a number of days after they stop being the latest, which keeps a versioned bucket's bill in check.

Applying it

aws s3api put-bucket-lifecycle-configuration --bucket market-docs-w8rjis --lifecycle-configuration file://lifecycle.json

aws s3api get-bucket-lifecycle-configuration --bucket market-docs-w8rjis

The put replaces every rule on the bucket, so one file holds them all. From then on S3 applies the rules once a day, with nothing for you to run.