Sessions Metrics

This page documents Prometheus metrics related to database session activity in Yellowbrick. These metrics provide insight into the number of active and idle sessions, as well as session-level memory consumption over time.

Purpose

Session metrics help operators and developers:

• Monitor how many sessions are active or idle across the system
• Identify long-idle sessions that may need to be closed (e.g. for resource hygiene)
• Analyze memory usage trends across sessions, including per-cluster patterns
• Tune connection pool behavior, session limits, or auto-idle disconnect logic

These metrics are critical for capacity planning, managing concurrency, and detecting orphaned client connections or resource exhaustion caused by excessive memory use.

Metrics

Name	Type	Freq	Labels	Description
yb_idle_session_count	gauge	30s	session_type	Number of idle database sessions
yb_idle_session_time_seconds_bucket	counter	30s	le, session_type	Cumulative count of sessions idle for <= bucket thresholds
yb_idle_session_time_seconds_count	counter	30s	session_type	Total number of idle sessions (for histogram)
yb_idle_session_time_seconds_sum	gauge	30s	session_type	Total sum of idle session times
yb_peak_session_memory_bytes_bucket	counter	1m	le, session_type, cluster, cluster_name	Cumulative count of sessions by peak memory usage (bytes)
yb_peak_session_memory_bytes_count	counter	1m	session_type, cluster, cluster_name	Total number of sessions contributing to peak memory usage
yb_peak_session_memory_bytes_sum	gauge	1m	session_type, cluster, cluster_name	Total peak memory usage across all sessions (in bytes)
yb_session_count	gauge	30s	session_type, cluster, cluster_name	Number of open database sessions