Origin Server Authentication plugin

重要

s3_auth plugin was replaced with origin_server_auth, which is 100% compatible with s3_auth plugin. For smooth transition, s3_auth plugin is available and provided as an alias of origin_server_auth plugin on this version of ATS. It is recommended that both the plugin name on your config files and the config file format for this plugin be updated as soon as possible. The alias and the support for s3_auth config format will be removed on a future release.

This is a plugin for Apache Traffic Server that provides support for Amazon S3 and Google Cloud Storage authentication features. This is useful if you for example want to use S3 as your origin server, yet want to avoid direct user access to the content.

Using the plugin

Using the plugin in a remap rule would be e.g.:

# remap.config

...  @plugin=origin_server_auth.so @pparam=--access_key @pparam=my-key \
                        @pparam=--secret_key @pparam=my-secret \
                        @pparam=--session_token @pparam=my-token \
                        @pparam=--virtual_host

Alternatively, you can store the access key and secret in an external configuration file, and point the remap rule(s) to it:

# remap.config

...  @plugin=origin_server_auth.so @pparam=--config @pparam=s3_auth_v2.config

Where s3.config could look like:

# s3_auth_v2.config

access_key=my-key
secret_key=my-secret
version=awsv2
virtual_host=yes

Both ways could be combined as well

Versions

New structure:

Value	Description
awsv2	AWS S3 version 2 (Default)
awsv4	AWS S3 version 4
gcpv1	Google Storage

Old structure (Deprecated):

Value	Description
2	AWS S3 version 2 (Default)
4	AWS S3 version 4

AWS Authentication version 4

The origin_server_auth plugin fully implements: AWS Signing Version 4 / Authorization Header / Transferring Payload in a Single Chunk / Unsigned Payload Option

Configuration options:

# Mandatory options
--access_key=<access_id>
--secret_key=<key>
--version=awsv4

# Optional
--session_token=<token>
--v4-include-headers=<comma-separated-list-of-headers-to-be-signed>
--v4-exclude-headers=<comma-separated-list-of-headers-not-to-be-signed>
--v4-region-map=region_map.config

If the following option is used then the options could be specified in a file:

--config=s3_auth_v4.config

The s3_auth_v4.config config file could look like this:

# s3_auth_v4.config

access_key=<access_id>
secret_key=<secret_key>
session_token=<token>
version=awsv4
v4-include-headers=<comma-separated-list-of-headers-to-be-signed>
v4-exclude-headers=<comma-separated-list-of-headers-not-to-be-signed>
v4-region-map=region_map.config

Where the region_map.config defines the entry-point hostname to region mapping i.e.:

# region_map.config

# "us-east-1"
s3.amazonaws.com                     : us-east-1
s3-external-1.amazonaws.com          : us-east-1
s3.dualstack.us-east-1.amazonaws.com : us-east-1

# us-west-1
s3-us-west-1.amazonaws.com           : us-west-1
s3.dualstack.us-west-1.amazonaws.com : us-west-1

# Default region if no entry-point matches:
: s3.amazonaws.com

If --v4-region-map is not specified the plugin defaults to the mapping defined in "Regions and Endpoints - S3"

According to Transferring Payload in a Single Chunk specification the CanonicalHeaders list must include the Host header, the Content-Type header if present in the request and all the x-amz-* headers so --v4-include-headers and --v4-exclude-headers do not impact those headers and they are always signed.

The Via and X-Forwarded-For headers are always excluded from the signature since they are meant to be changed by the proxies and signing them could lead to invalidation of the signature.

If --v4-include-headers is not specified all headers except those specified in --v4-exclude-headers will be signed.

If --v4-include-headers is specified only the headers specified will be signed except those specified in --v4-exclude-headers

AWS Authentication version 2

For more details on the S3 auth version 2 , see: Signing and Authenticating REST Requests

There are 4 plugin configuration options for version 2:

--access_key    <access_id>
--secret_key    <secret_key>
--virtual_host
--config        <config file>
--version=awsv2

This is a pretty bare bone start for the S3 services, it is missing a number of features:

• It does not do UTF8 encoding (as required)

• It does not deal with canonicalization of AMZ headers.

• It does not handle POST requests (but do we need to ?)

• It does not incorporate query parameters.

• It does not support session tokens.

Contributions to any of these would be appreciated.

Google Cloud Storage

Configuration options:

# Mandatory options
--session_token=<session_token>
--version=gcpv1

If the following option is used then the options could be specified in a file:

@plugin=origin_server_auth.so @pparam=--config @pparam=gcp_auth.config

The gcp_auth.config config file could look like this:

# gcp_auth.config

session_token=<access_id>
version=gcpv1

Retrying config loading

If the specified configuration file cannot be opened or is missing required options, ATS will attempt to reload the file repeatedly with exponential backoff.

If the configuration file includes an expiration parameter and the file has exceeded its expiration time, ATS will retry loading the file every minute for a duration of 10 minutes. After 10 minutes, the file must be manually reloaded.