Enable Thumbor to access S3 resources using AWS Managed Policies #35

tschaffter · 2024-09-20T17:48:49Z

Thank you for developing and maintaining this Thumbor Docker image! It's been a great help to us in our project, and we appreciate all the effort you've put into it.

We encountered an issue while attempting to provide the Docker container on AWS ECS access to an S3 bucket using an AWS managed policy instead of passing AWS credentials directly. When we try to access an image, Thumbor throws an InvalidAccessKeyId error. This happens when attempting to fetch an image from within the container.

Error Details:

2024-09-20 17:24:52 thumbor:ERROR ERROR: Traceback (most recent call last):
File "/usr/local/lib/python3.10/site-packages/thumbor/handlers/__init__.py", line 212, in get_image
    result = await self._fetch(self.context.request.image_url)
File "/usr/local/lib/python3.10/site-packages/thumbor/handlers/__init__.py", line 884, in _fetch
    loader_result = await self.context.modules.loader.load(
File "/usr/local/lib/python3.10/site-packages/thumbor_aws/loader.py", line 89, in load
    status_code, body, last_modified = await client.get_data(
File "/usr/local/lib/python3.10/site-packages/thumbor_aws/s3_client.py", line 143, in get_data
    response = await client.get_object(Bucket=bucket, Key=path)
File "/usr/local/lib/python3.10/site-packages/aiobotocore/client.py", line 371, in _make_api_call
    raise error_class(parsed_response, operation_name)
botocore.exceptions.ClientError: An error occurred (InvalidAccessKeyId) when calling the GetObject operation: The AWS Access Key Id you provided does not exist in our records.

Steps to Reproduce:

Configure Thumbor to use an AWS managed policy for S3 access.
Try to access an image hosted in the S3 bucket.
Observe the InvalidAccessKeyId error in the logs.

Expected Behavior:
Thumbor should successfully access the S3 bucket using the permissions granted by the managed policy, without the need for hardcoded AWS credentials.

Actual Behavior:
Thumbor fails to access the S3 bucket and throws an InvalidAccessKeyId error.

Suggested Solution:
It seems that Thumbor might not be properly configured to use IAM roles or managed policies. We suggest updating the Thumbor configuration to fully support AWS managed policies for S3 access. This could involve:

Ensuring that the AWS SDK within Thumbor is correctly set up to use the IAM role or managed policy assigned to the ECS task/container.
Verifying that Thumbor is not defaulting to looking for static credentials in the environment or configuration files.

Environment:

Thumbor Version: beeyev/thumbor-s3:7.4-alpine
Deployment: ECS Fargate with IAM roles

Additional Context:
This issue prevents us from using the recommended security practices for AWS deployments, where sensitive credentials should not be hardcoded or passed explicitly.

We would appreciate any guidance on how to configure Thumbor to work with AWS managed policies, or if this could be implemented as a feature.

The text was updated successfully, but these errors were encountered:

tschaffter · 2024-09-20T18:00:35Z

Similar issues submitted to other Thumbor Docker images:

tschaffter · 2024-09-20T21:10:50Z

I can successfully download an image in the S3 bucket with the AWS CLI from inside the Thumbor container. This command is run right after installing the AWS CLI client, without configuring any AWS profile or credentials thanks to the AWS managed policy that we have set in place:

/app # aws s3 cp s3://BUCKET_NAME/img/triforce.png triforce.png
download: s3://BUCKET_NAME/img/triforce.png to ./triforce.png

tschaffter · 2024-09-25T18:33:37Z

@beeyev There is one difference between the config example provided by thumbor_aws and the config file generated by the Docker image. The former uses None values while the latter specifies 'None'.

thumbor_aws uses None: https://github.com/thumbor/thumbor-aws/tree/main?tab=readme-ov-file#loader

This image generates a config file that includes 'None': from the actual config file inside the container: /usr/local/etc/thumbor.conf:

## Secret access key for S3 Loader.
## Defaults to: None
AWS_LOADER_S3_SECRET_ACCESS_KEY = 'None'

## Access key ID for S3 Loader.
## Defaults to: None
AWS_LOADER_S3_ACCESS_KEY_ID = 'None'

I think that thumbor_aws evaluates the value as the string "None" and not as the None value.

beeyev · 2024-09-25T18:44:41Z

Thank you, I will check it!

tschaffter · 2024-09-25T21:48:00Z

I can now confirm that specifying None instead of 'None' solved the issue.

beeyev · 2024-09-28T15:25:27Z

@tschaffter
I built a new test docker image with possible fix ghcr.io/beeyev/thumbor-s3:debian-test
could you please check if it works?

tschaffter · 2024-09-30T14:44:06Z

I will give it a try today. Thanks!

tschaffter · 2024-09-30T16:18:15Z

@beeyev I get this error when stepping into the container and fetching an image with wget.

Event Time	Message
2024-09-30T16:04:34.403Z	2024-09-30 16:04:34 thumbor:ERROR ERROR: Traceback (most recent call last):
2024-09-30T16:04:34.403Z	File "/usr/local/lib/python3.12/site-packages/thumbor/handlers/init.py", line 214, in get_image
2024-09-30T16:04:34.403Z	result = await self._fetch(self.context.request.image_url)
2024-09-30T16:04:34.403Z	^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2024-09-30T16:04:34.403Z	File "/usr/local/lib/python3.12/site-packages/thumbor/handlers/init.py", line 916, in _fetch
2024-09-30T16:04:34.403Z	loader_result = await self.context.modules.loader.load(
2024-09-30T16:04:34.403Z	^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2024-09-30T16:04:34.403Z	File "/usr/local/lib/python3.12/site-packages/thumbor_aws/loader.py", line 89, in load
2024-09-30T16:04:34.403Z	status_code, body, last_modified = await client.get_data(
2024-09-30T16:04:34.403Z	^^^^^^^^^^^^^^^^^^^^^^
2024-09-30T16:04:34.403Z	File "/usr/local/lib/python3.12/site-packages/thumbor_aws/s3_client.py", line 141, in get_data
2024-09-30T16:04:34.403Z	async with self.get_client() as client:
2024-09-30T16:04:34.403Z	^^^^^^^^^^^^^^^^^
2024-09-30T16:04:34.403Z	File "/usr/local/lib/python3.12/site-packages/thumbor_aws/s3_client.py", line 95, in get_client
2024-09-30T16:04:34.403Z	region_name=self.region_name,
2024-09-30T16:04:34.403Z	^^^^^^^^^^^^^^^^
2024-09-30T16:04:34.403Z	File "/usr/local/lib/python3.12/site-packages/thumbor_aws/s3_client.py", line 46, in region_name
2024-09-30T16:04:34.403Z	"region_name", self.config.AWS_STORAGE_REGION_NAME
2024-09-30T16:04:34.403Z	^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2024-09-30T16:04:34.403Z	File "/usr/local/lib/python3.12/site-packages/derpconf/config.py", line 225, in getattr
2024-09-30T16:04:34.403Z	raise AttributeError(name)
2024-09-30T16:04:34.403Z	AttributeError: AWS_STORAGE_REGION_NAME. Did you mean: 'AWS_LOADER_REGION_NAME'?
2024-09-30T16:04:34.403Z	2024-09-30 16:04:34 thumbor:ERROR [BaseHandler] get_image failed for url `triforce.png`. error: `AWS_STORAGE_REGION_NAME`
2024-09-30T16:04:34.403Z	2024-09-30 16:04:34 tornado.access:ERROR 500 GET /unsafe/600x600/triforce.png (127.0.0.1) 4.49ms

I do not specify the environment variable AWS_STORAGE_REGION_NAME when starting the container.

# echo $AWS_STORAGE_REGION_NAME

This Key-Value is not specified in the config file because this PR only prints some KVs if their key is specified.

$ cat /usr/local/etc/thumbor.conf
...

################################# AWS Storage ##################################
# Documentation: https://github.com/thumbor/thumbor-aws#storage

## Region where thumbor's objects are going to be stored.
## Defaults to: 'us-east-1'


## S3 Bucket where thumbor's objects are going to be stored.
## Defaults to: 'thumbor'

beeyev · 2024-09-30T17:10:50Z

What configuration do you use? Can you provide the values please

tschaffter · 2024-09-30T17:14:42Z

Here are the environment variables passed to the Thumbor container deployed with AWS CDK (dev deployment):

thumbor_props = ServiceProps(
    "openchallenges-thumbor",
    8889,
    512,
    "ghcr.io/beeyev/thumbor-s3:debian-test",
    {
        "LOG_LEVEL": "info",
        "PORT": "8889",
        "LOADER": "thumbor_aws.loader",
        "AWS_LOADER_REGION_NAME": "us-east-1",
        "AWS_LOADER_BUCKET_NAME": bucket_stack.openchallenges_img_bucket.bucket_name,
        "AWS_LOADER_S3_ENDPOINT_URL": "http://s3.us-east-1.amazonaws.com",
        "AWS_LOADER_ROOT_PATH": "img",
        "STORAGE": "thumbor.storages.file_storage",
        "FILE_STORAGE_ROOT_PATH": "/data/storage",
        "RESULT_STORAGE": "thumbor.result_storages.file_storage",
        "RESULT_STORAGE_FILE_STORAGE_ROOT_PATH": "/data/result_storage",
        "RESULT_STORAGE_STORES_UNSAFE": "True",
        "RESULT_STORAGE_EXPIRATION_SECONDS": "2629746",
        "SECURITY_KEY": secrets["SECURITY_KEY"],
        "ALLOW_UNSAFE_URL": "True",
        "QUALITY": "100",
        "MAX_AGE": "86400",
        "AUTO_PNG_TO_JPG": "True",
        "HTTP_LOADER_VALIDATE_CERTS": "False",
    },
)

beeyev · 2024-09-30T17:41:27Z

@tschaffter
I just updated the docker image ghcr.io/beeyev/thumbor-s3:debian-test
Could you please pull the update and try again?

tschaffter · 2024-09-30T18:30:39Z

@beeyev The latest image works!

beeyev · 2024-09-30T18:41:52Z

Amazing! I will prepare the new release tomorrow

tschaffter · 2024-09-30T18:48:50Z

Thanks so much for fixing that so quickly.

Config update, fix issue #35 (#36)

beeyev · 2024-10-01T10:13:07Z

@tschaffter I updated the docker image with the fix.
Thank you for your help solving this issue 👍

tschaffter · 2024-10-01T15:23:17Z

@beeyev Could you please update the image beeyev/thumbor-s3:7.7-alpine too (last updated 10 months ago)? I need SVG to PNG conversion, which is not available in the slim version.

https://hub.docker.com/r/beeyev/thumbor-s3/tags
https://github.com/beeyev/thumbor-s3-docker/pkgs/container/thumbor-s3

beeyev · 2024-10-01T17:26:41Z

@tschaffter I deprecated this image, because full-featured thumbor instance can't work stable on alpine linux.
You need to use debian if you want all features, or you can use slim-alpine if you are fine with the basic functionality.

beeyev/thumbor-s3:debian
beeyev/thumbor-s3:slim-alpine
beeyev/thumbor-s3:7-debian
beeyev/thumbor-s3:7-slim-alpine
beeyev/thumbor-s3:7.7-debian
beeyev/thumbor-s3:7.7-slim-alpine

tschaffter mentioned this issue Sep 20, 2024

chore(openchallenges): update Thumbor Docker image Sage-Bionetworks/sage-monorepo#2838

Merged

tschaffter mentioned this issue Sep 25, 2024

fix(openchallenges): patch the entry point script of Thumbor to replace 'None' by None (IT-3908) Sage-Bionetworks/sage-monorepo#2856

Merged

beeyev mentioned this issue Sep 27, 2024

Config update, fix issue #35 #36

Merged

beeyev added a commit that referenced this issue Oct 1, 2024

Config update, fix issue #35 (#36)

902e2c0

Config update, fix issue #35 (#36)

tschaffter mentioned this issue Oct 2, 2024

fix(openchallenges): update the Thumbor image to support AWS managed policies (ARCH-302) Sage-Bionetworks/sage-monorepo#2870

Merged

beeyev closed this as completed Oct 2, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Enable Thumbor to access S3 resources using AWS Managed Policies #35

Enable Thumbor to access S3 resources using AWS Managed Policies #35

tschaffter commented Sep 20, 2024

tschaffter commented Sep 20, 2024 •

edited

Loading

tschaffter commented Sep 20, 2024 •

edited

Loading

tschaffter commented Sep 25, 2024

beeyev commented Sep 25, 2024

tschaffter commented Sep 25, 2024

beeyev commented Sep 28, 2024

tschaffter commented Sep 30, 2024

tschaffter commented Sep 30, 2024 •

edited

Loading

beeyev commented Sep 30, 2024

tschaffter commented Sep 30, 2024 •

edited

Loading

beeyev commented Sep 30, 2024

tschaffter commented Sep 30, 2024

beeyev commented Sep 30, 2024

tschaffter commented Sep 30, 2024

beeyev commented Oct 1, 2024

tschaffter commented Oct 1, 2024

beeyev commented Oct 1, 2024

Enable Thumbor to access S3 resources using AWS Managed Policies #35

Enable Thumbor to access S3 resources using AWS Managed Policies #35

Comments

tschaffter commented Sep 20, 2024

tschaffter commented Sep 20, 2024 • edited Loading

tschaffter commented Sep 20, 2024 • edited Loading

tschaffter commented Sep 25, 2024

beeyev commented Sep 25, 2024

tschaffter commented Sep 25, 2024

beeyev commented Sep 28, 2024

tschaffter commented Sep 30, 2024

tschaffter commented Sep 30, 2024 • edited Loading

beeyev commented Sep 30, 2024

tschaffter commented Sep 30, 2024 • edited Loading

beeyev commented Sep 30, 2024

tschaffter commented Sep 30, 2024

beeyev commented Sep 30, 2024

tschaffter commented Sep 30, 2024

beeyev commented Oct 1, 2024

tschaffter commented Oct 1, 2024

beeyev commented Oct 1, 2024

tschaffter commented Sep 20, 2024 •

edited

Loading

tschaffter commented Sep 20, 2024 •

edited

Loading

tschaffter commented Sep 30, 2024 •

edited

Loading

tschaffter commented Sep 30, 2024 •

edited

Loading