LangChain

Error Rate Masking Latency Degradation

warning
reliabilityUpdated Feb 12, 2026

High langchain_request_error or langchain_chain_error rates can suppress latency metrics (fast-failing requests skew averages downward), hiding underlying performance issues that affect successful requests.

How to detect:

Alert when langchain_request_error or langchain_chain_error exceeds threshold (e.g., >5% of langchain_chain_run) AND p95 langchain_request_time remains stable or decreases. This pattern suggests errors are terminating quickly, not reflecting true user-facing latency.

Recommended action:

Separate success vs. error latency distributions in dashboards. Investigate error root causes (rate limits, timeouts, model unavailability) and monitor success-only p95/p99 latency. Use LangSmith Alerts to set thresholds on both error rate and success latency independently.