Skip to content

datasets-server.huggingface.co returning 503 on every endpoint (whole host down, not one route) #8331

Description

@ChristianDenniss

Describe the bug

The entire datasets-server.huggingface.co backend (the dataset-viewer API) is
returning 503 Service Temporarily Unavailable on every route I've tried, not
just one endpoint. It reproduces on public datasets with no auth token, so it
doesn't appear to be account- or dataset-specific. huggingface.co itself and
huggingface.co/api are unaffected, so this looks isolated to the dataset-viewer
service's backend/load balancer.

Steps to reproduce the bug

Run any of the following (curl, no auth needed):

curl -i "https://datasets-server.huggingface.co/"
curl -i "https://datasets-server.huggingface.co/info?dataset=squad&config=plain_text"
curl -i "https://datasets-server.huggingface.co/splits?dataset=squad"
curl -i "https://datasets-server.huggingface.co/rows?dataset=squad&config=plain_text&split=validation&offset=0&length=1"
curl -i "https://datasets-server.huggingface.co/parquet?dataset=squad"
curl -i "https://datasets-server.huggingface.co/valid"
curl -i "https://datasets-server.huggingface.co/is-valid?dataset=squad"
curl -i "https://datasets-server.huggingface.co/size?dataset=squad"
curl -i "https://datasets-server.huggingface.co/statistics?dataset=squad&config=plain_text&split=validation"

All of the above return an identical response:

HTTP/1.1 503 Service Temporarily Unavailable
Content-Type: text/html
Content-Length: 162
Connection: keep-alive
Server: awselb/2.0
X-Cache: Error from cloudfront
Via: 1.1 d9d0b19761149aebd7234df3fac341aa.cloudfront.net (CloudFront)
X-Amz-Cf-Pop: YUL62-P1

Run any of the following (curl, no auth needed):

```bash
curl -i "https://datasets-server.huggingface.co/"
curl -i "https://datasets-server.huggingface.co/info?dataset=squad&config=plain_text"
curl -i "https://datasets-server.huggingface.co/splits?dataset=squad"
curl -i "https://datasets-server.huggingface.co/rows?dataset=squad&config=plain_text&split=validation&offset=0&length=1"
curl -i "https://datasets-server.huggingface.co/parquet?dataset=squad"
curl -i "https://datasets-server.huggingface.co/valid"
curl -i "https://datasets-server.huggingface.co/is-valid?dataset=squad"
curl -i "https://datasets-server.huggingface.co/size?dataset=squad"
curl -i "https://datasets-server.huggingface.co/statistics?dataset=squad&config=plain_text&split=validation"

All of the above return an identical response

Expected behavior

datasets-server.huggingface.co should respond normally (200, or an appropriate
4xx for a bad dataset/config/split) rather than a blanket 503 with an
awselb/cloudfront error page across every route.

Environment info

N/A - this is a direct HTTP-level issue against the datasets-server.huggingface.co
backend itself (via curl), not something surfaced through the datasets Python
library, so there's no local datasets/python/platform version relevant here.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions