Dataset edit page returns HTTP 500 / ServiceError - eu-west-1

We are experiencing a server-side issue in Amazon QuickSight when opening several datasets in edit mode.

Summary:
When we open a dataset to edit it, the page fails to load and shows “Something went wrong. Try again.” (see attached screenshots). This affects multiple datasets, not just one.

Key point — this is not a client-side problem:

  • The datasets are healthy: scheduled refreshes complete successfully and all dashboards built on top of them load and work correctly.
  • The failure only happens when opening the dataset EDIT view.
  • We reproduced it in Chrome and Firefox, both in a clean incognito/private window with extensions disabled and cache cleared.

Browser console shows a server error:

  • “Failed to load resource: the server responded with a status of 500”
  • Followed by “ServiceError: null” thrown from the dataset-management bundle
  • Call stack: RetryableFetchAppServiceClient.fetch → Object.queryFn (dataset-management-v…js)
  • The underlying failing request is a GET to:
    https://eu-west-1.quicksight.aws.amazon.com/sn/account/

Environment:

  • Region: eu-west-1
  • Example dataset: Quick/Analisi/Clienti/Accounts

Since this is returning an HTTP 500 from the QuickSight service itself, could you please check the service-side logs for these dataset edit requests and let us know the root cause and resolution? Happy to provide the exact request ID, dataset ARNs, and a UTC timestamp window if that helps you locate the failing calls.

&column-identifier-mode=ID_AND_NAME enpoint without this option return correct JSON

Hi @Massimo and welcome to the community!

Thanks for providing so much detail in your post and for confirming the steps you’ve already taken/important and relevant information. At this point, I believe your best bet is to open a support ticket to get further assistance from the AWS team. (Case management - AWS Support).

Hope this gets resolved and feel free to keep us updated!

Update — root cause found:

I’ve isolated a reproducible trigger. The 500 on DescribeDataSet with column-identifier-mode=ID_AND_NAME is caused by column descriptions on a SPICE dataset that is then used as a source in a JOIN inside another dataset.

Minimal repro:

  1. Take a SPICE dataset and populate the per-column descriptions.
  2. Use that dataset as a source table in a JOIN inside another dataset.
  3. Open the parent (joined) dataset in the edit view → “Something went wrong. Try again.” (DescribeDataSet with ID_AND_NAME returns HTTP 500, body { "message": null }).
  4. Clear all column descriptions on the source SPICE dataset → the parent dataset opens fine, error gone.

So the trigger is column descriptions on a joined source dataset not being handled correctly by the ID_AND_NAME resolution path. Removing the descriptions is the workaround. These datasets, joins and descriptions had worked for over a year with no changes on our side, so this looks like a recent service-side regression.

If delete description columns not error, the child dataset edit is OK