Milvus

Strong Consistency tSafe Wait Penalty

warning
latencyUpdated Oct 2, 2025

When strict (Strong) consistency is enabled, queries must wait for tSafe (time-safe) synchronization across all nodes, adding significant latency overhead especially in distributed deployments or during high write throughput periods.

How to detect:

Monitor wait tSafe latency metrics, particularly for Strong consistency queries. Alert when tSafe wait time exceeds 50ms or represents >30% of total query latency. Correlate with rootcoord sync timetick time and proxy tt_lag metrics to identify synchronization delays.

Recommended action:

Evaluate whether Strong consistency is required for all queries; downgrade to Bounded or Eventually consistent reads where stale reads are acceptable (e.g., analytics, recommendation systems). Optimize write throughput to reduce tSafe lag. Monitor and tune rootcoord.dmlChannelNum and common.gracefulTime parameters.