Request Backlog Indicates Processing Bottleneck
warninglatencyUpdated Feb 6, 2026
When outstanding_requests grows above zero consistently, ZooKeeper cannot keep up with incoming request load. This leads to increased latency and potential client session timeouts.
Sources
How to Monitor Zookeeper Ensemble Health with the Collectoroneuptime.com
How to Monitor ZooKeeper: Key Metrics & Best Tools in 2023sematext.com
Configuring and Monitoring Zookeeper with OpenTelemetry Receiveropenobserve.ai
End-to-End Guide: Configuring and Monitoring Zookeeper with OpenTelemetry Receiveropenobserve.ai
How to Configure the Zookeeper Receiver in the OpenTelemetry ...oneuptime.com
Technologies:
How to detect:
Monitor zookeeper.outstanding_requests - it should remain at or near zero. Alert when it exceeds 10 requests for more than 1 minute, or when average latency exceeds 50ms while outstanding requests are growing.
Recommended action:
Investigate the root cause: check disk I/O latency (especially transaction log disk), examine JVM GC pause times, verify CPU utilization is below 70%, and review network latency between clients and servers. Consider scaling horizontally by adding read-only observers if read load is high.