DMS periodic instability on az-eastus-1, 5XX responses

Update timeline

1. investigating May 21, 2026, 07:10 PM UTC

   We are currently investigating periodic instability affecting Data Modeling Storage (DMS) services on the az-eastus-1 cluster. Users may experience intermittent 5XX errors when accessing Data Modeling endpoints. Our engineering team is actively investigating.

2. investigating May 21, 2026, 07:55 PM UTC

   Problem: The datamodelstorage service in az-eastus-1 is experiencing periodic spikes to 100% 5xx error rates, resulting in major outages for users. Impact: Multiple customers are receiving repeated 503 errors. All users on this cluster are likely affected. Sev A declared.

3. identified May 21, 2026, 08:20 PM UTC

   The engineers have identified the root cause and are working on a solution.

4. identified May 21, 2026, 08:26 PM UTC

   Engineers have identified a mitigation action and are working on a temporary solution, as well as monitoring the situation.

5. monitoring May 21, 2026, 08:52 PM UTC

   Cognite has reviewed deployment history, logs, and network dependencies. An internal process was found to be blocking operations. Cognite has stopped that process. The rates on the 500 errors has dropped, and the responsible engineering teams will resume investigation tomorrow.

6. monitoring May 22, 2026, 02:37 PM UTC

   Cognite has deployed a fix for the issue causing intermittent 5xx errors on the Data Modeling Service (DMS) in the az-eastus-1 cluster. All related SLO alerts have cleared and background tasks have resumed normal operation. We are continuing to observe the cluster closely before resolving the incident. We will post a further update once monitoring confirms full stability.

7. resolved May 22, 2026, 05:56 PM UTC

   The issue affecting Data Modeling Storage (DMS) services on the az-eastus-1 cluster has been resolved. All services are now operating normally.