Skip to content

Managed Grafana SSO login fails with 500 "could not refresh user information"

0

Workspace: g-faba4f9f70 (ap-southeast-2) Account: 238663102864 Grafana version: 12.4.3 Identity Center instance: ssoins-82598c0b7b7a9125 (org instance, management account 992382375457)

All SSO logins to the workspace began failing on 2026-10-15 21:11:13 UTC. Last successful login was 2026-10-14 05:07:17 UTC.

CloudTrail shows every login attempt returning: eventName: login_auth_sso errorCode: 500 errorMessage: [sso.auth] could not refresh user information: internal server error

The failures began before any configuration changes were made on our side.

Already verified / attempted with no change:

  • Workspace status Active, authentication provider AWS_SSO enabled
  • Both users present and active in Identity Center, IDs unchanged (confirmed against the AuthId values in the last successful login events)
  • All Identity Center groups the users belong to still exist
  • Users can sign in to the AWS access portal normally
  • Users assigned to the workspace with roles (Admin / Editor)
  • Users assigned to the Amazon Grafana application in Identity Center
  • Deleted and recreated the Identity Center authentication configuration on the workspace, then re-assigned users - same error
  • Tested in a clean incognito session

The 500 suggests a server-side failure in the SSO user-info refresh rather than an authorisation problem. Has anyone seen this, or is there a known issue with 12.4.3?

1 Answer
1

I couldn't find any publicly reported known issue for this error or for Grafana 12.4.3 on Amazon Managed Grafana. Your reading is right: an HTTP 500 during the user-info refresh is a server-side failure, not an authorization denial. Two things can cause it: a problem on the Managed Grafana service side, or something in Identity Center that stops Grafana from reading user and group data. I'd pursue both in parallel.

1. Open a support case now

You've ruled out everything you can change in the workspace itself, so AWS has to look at the backend. Open the case under Amazon Managed Grafana at "production system down" severity, and include:

  • the workspace ID, region and Grafana version
  • the exact time of the first failure and of the last successful login
  • the requestID / eventID values from a few failing login_auth_sso CloudTrail events
  • a list of everything you've already tried (your list is good)

2. Check the Identity Center side, in the management account

Grafana fetches user and group information from Identity Center when someone logs in. Changes at the org level can break that even when the access portal still works.

  • Customer managed KMS key for Identity Center. If Identity Center was switched to a customer managed key around 15 October, or that key's policy was changed or the key disabled, AWS managed applications can fail to read identity data. Look in CloudTrail for KMS Decrypt/GenerateDataKey errors involving the Identity Center key. If you find them, add the application-access statements AWS documents to the key policy.
  • CloudTrail around 2026-10-15 21:11 UTC. Look for any of these:
    • changes to the identity source or SCIM provisioning, e.g. a resync or attribute mapping change
    • group membership changes that affect these users
    • failed identitystore:* or sso:* calls whose invoking service is grafana.amazonaws.com
  • SCPs and RCPs. Check whether a policy was attached or changed that denies Identity Center or Identity Store actions for the workspace account or the management account.
  • Attribute changes from the identity provider. If you use an external IdP with SCIM, check whether email, username or display name changed for these users. A malformed or missing attribute can break a user-info refresh even though the user IDs stay the same.

3. Find out whether the problem is the workspace or Identity Center

Create a throwaway workspace in the same account and region with the same Identity Center instance, assign one of the users, and try logging in.

  • If the test workspace also fails: the cause is Identity Center, the KMS key, an SCP, or service-wide.
  • If the test workspace works: the cause is specific to g-faba4f9f70, for example a backend change or upgrade to that workspace. Tell Support, because it narrows their search a lot.

To test whether the Grafana version matters, you can also create the test workspace on an older version such as 10.4.

4. Temporary access while you wait

  • Managed Grafana lets you enable SAML authentication alongside Identity Center. You could add a SAML app, using your external IdP or a custom SAML application in Identity Center, to get admins back in.
  • Service account tokens keep automation and API access working, because they don't depend on SSO login.
EXPERT

answered a month ago

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.