Collector Queue Saturation Leading to Trace Loss
criticalResource ContentionUpdated Feb 7, 2026
When Jaeger collector's internal queue exceeds 70-80% capacity, spans begin queueing and risk being dropped, resulting in incomplete traces and data loss.
Sources
How to detect:
Monitor collector queue length approaching capacity. Queue utilization above 70% indicates backpressure; above 80% is critical. Also watch for divergence between spans received and traces saved.
Recommended action:
Scale collector horizontally by adding more instances. Check storage backend write latency and CPU/memory limits on collectors. If using Kafka, verify ingester throughput. Alert when queue utilization crosses 80%.