Is Cognite down?

We are currently investigating an issue impacting the Time Series API in the statnett cluster. Current Impact: • Timeseries Metadata Read and Subscription Read are currently experiencing a total outage. • Datapoint ingestion and Datapoint read are partially impacted with elevated error rates and latency Our engineering team has identified a certificate validation issue connecting to internal caching services and is actively working on refreshing credentials to restore affected pods and services. We will provide further updates as progress is made.

We identified a certificate configuration issue that prevented the Time Series API in the statnett cluster from connecting to an internal caching service. This caused an outage for time-series metadata and subscription reads, and partially affected datapoint ingestion and reads. Service has been restored by correcting the certificate configuration

We are monitoring the Cognite Data Fusion API in the statnett cluster to ensure the service remains stable following the restoration.

This incident has been resolved. The Time Series API in the Statnett cluster has been restored and is operating normally. All affected services are functioning as expected.

Newly uploaded files in the Tokyo region (az-tyo-gp-001) may remain marked as not uploaded and unavailable for longer than usual. This affects downloads, viewing newly uploaded photos in InField, and workflows that wait for upload completion. File contents already in storage are unchanged. We are continuing to investigate the full scope of the impact.

Newly uploaded files in the Tokyo region (az-tyo-gp-001) may take longer than usual to become available. This affects downloads, viewing newly uploaded photos in InField, and workflows that depend on upload completion. Downloads and workflows that wait for a file to finish uploading are affected. File contents already in storage are unchanged. We have identified an issue with upload-status processing and are working on a fix.

Monitoring: We've applied a fix for the issue where files uploaded via uploadUrl weren't showing as uploaded and couldn't be accessed. File upload processing has recovered in the affected clusters and customers have confirmed their files can be accessed again. We are now monitoring to make sure the remaining processing backlog clears and upload status updates normally in all clusters. We'll post another update once we've confirmed everything is back to normal.

Files uploaded via uploadUrl in az-tyo-gp-001 are now available as expected, including in InField.

We are investigating reports of service degradation and disruption affecting the 3D service in the az-tyo-002 cluster. 3D models and related API endpoints may be unavailable. Our engineering team is currently investigating the issue

The 3D service in the az-tyo-002 cluster is experiencing service disruptions due to repeated service restarts during startup health checks. The engineering team has identified the mitigation and is currently deploying an adjustment to the startup probe delays to stabilize the pods and restore service availability

The fix increased the health probe startup delay to 60 seconds. The 3D service in az-tyo-002 is back online, and pods appear to have stopped restarting. The engineering team is now Monitoring and checking the stability while awaiting customer confirmation.

The issue affecting 3D services in az-tyo-002 has been fixed by increasing the health probe startup delay. Service has been restored, and is now resolved.

We are investigating elevated errors affecting Atlas AI agent chat across multiple regions. Users may experience failed chat requests. Initial findings suggest upstream timeouts from Azure OpenAI in the East US 2 region may be contributing to the issue, including requests originating from other regions. We are continuing to investigate the cause and impact.

Upstream provider Microsoft has confirmed that the underlying service-side timeout issues in Azure OpenAI (East US 2) are fully resolved. Customer-facing mitigation remains in place and all Atlas AI agent chat metrics, traffic and availability have returned to normal operating baselines across all clusters.

We experienced an issue with the Time Series API that caused temporary service disruptions across select clusters: • az-arn-001: Time series datapoint ingestion (ingesting and deleting data points) experienced HTTP 503 errors and unavailability between 13:49 and 14:02 CEST. • az-sin-sp-001: Time series metadata endpoints (listing, searching, and creating classic time series) experienced a brief period of errors between 14:02 and 14:05 CEST. Datapoint ingestion remained operational.

Our engineering team identified the regression and completed a rollback to restore service. All services have recovered and are operating normally. We are actively monitoring the systems to ensure ongoing stability

All services have recovered and are operating normally.