From approximately 1716 UTC to 1751 UTC the LangSmith UI and Run Ingestion experienced degraded performance.
During this time, all runs submitted to LangSmith were successfully ingested but may have been delayed in appearing in the LangSmith UI.
We identified a set of inefficient queries related to filtering in the Threads view for Projects that, when run against very large projects, were starving our production database of resources. We have shipped optimized versions of those queries that significantly improve performance and reduce CPU and memory utilization and have not seen a recurrence.
Resolved
From approximately 1716 UTC to 1751 UTC the LangSmith UI and Run Ingestion experienced degraded performance.
During this time, all runs submitted to LangSmith were successfully ingested but may have been delayed in appearing in the LangSmith UI.
We identified a set of inefficient queries related to filtering in the Threads view for Projects that, when run against very large projects, were starving our production database of resources. We have shipped optimized versions of those queries that significantly improve performance and reduce CPU and memory utilization and have not seen a recurrence.
Monitoring
Latency has returned to historic norms. We are monitoring and continuing to investigate the root cause of the slowdown.
Investigating
We have received reports of increased latency in traces appearing in LangSmith and slow loading or errors in the LangSmith Frontend and are actively investigating the issue.
Looking for the EU status page? Find it here: https://eu.status.smith.langchain.com