I am writing this S3 bucket lifecycle rule:

 lifecycle_rules = [
    {
      id      = "log"
      enabled = true
      prefix  = "log/"

      tags = {
        "rule"      = "log"
        "autoclean" = "true"
      }
      expiration = {
        days = 365
      }
    }
  ]

An ASCII mock diagram of the bucket is provided below: Logs are dynamically written everyday in this bucket. Some in the folder logs/, some in the subfolder failed_logs/ and some on the root folder of the bucket.

BUCKET/
|
  nonlogfile.exe
  log_file_that_i_DONT_wanna_delete.log
  logs/ 
      |
       file1     (Created 2021-01-01)
       file2     (Created 2021-01-02)
       failed_logs/
                  |
                   file3 (Created 2021-01-03) 
                   file4 (Created 2021-01-04)

My target is to expire everyfile in the log/ folder (and its subfolder failed_logs after 365 days.

`file1` would expire at 2022-01-01
`file2` would expire at 2022-01-02
`file3` would expire at 2022-01-03
`file4` would expire at 2022-01-04

whereas

nonlogfile.exe
log_file_that_i_DONT_wanna_delete.log

would never expire.

Terraform's aws_s3_bucket documentation states that:

The lifecycle_rule object supports the following:

prefix - (Optional) Object key prefix identifying one or more objects to which the rule applies.

tags - (Optional) Specifies object tags key and value.

• I don't understand the prefix attribute
• My files are not tagged

How can I target them as explained above, based on their location on the bucket?

CodePudding user response：

S3 Structure Explained

The prefix attribute is simply the path to your s3 object minus the bucket name. If you had the following S3 "directory" s3://ec-integration-files/logs/ the prefix value would simply be logs/.

The attribute is named prefix as opposed to path because while s3 looks like a directory tree structure technically it is flat storage. However the Management Console displays the objects as if they are stored in a directory structure, and it uses the prefixes to do this.

One other point of clarification, you can have two objects at the root ( or at any depth ) named logs and logs/. One of these will be displayed as a directory in the management console and the other as a file object, but they are really just two objects with different names