The Buffer Cache Hit Ratio alert warns when Oracle serves too few cacheable reads from memory after enough activity has been sampled.
Mini DBA evaluates the percentage of Oracle cacheable reads served from the buffer cache after enough logical-read activity has been sampled. Missing or immature samples are ignored, so a new, idle, or not-yet-populated instance does not alert as 0%.
A low hit ratio is most useful when it appears with physical read pressure. In that context it can indicate excessive physical I/O, a working set larger than the SGA, or inefficient SQL that performs avoidable scans. Mini DBA evaluates this alert once a minute, and duration prevents a single short spike from immediately becoming an incident.
This alert helps detect memory pressure and cache efficiency problems. Low cache effectiveness or high memory pressure increases physical I/O, query latency, and the chance that a workload will become unstable during busy periods.
Enable it once the instance has a normal workload baseline. It is especially useful on busy OLTP, reporting, and integration systems where resource pressure can quickly become user-visible.
Threshold meaning: %. Major threshold: 90 %. Minor threshold: 95 %. Comparison direction: "under". Duration is used, so prefer requiring the condition to persist before paging people for transient spikes. Use higher thresholds on batch-heavy, development, or intentionally bursty systems where brief pressure is expected. Use lower thresholds on latency-sensitive production systems, small instances with little headroom, and services with strict recovery or availability commitments.
Check memory configuration, cache hit indicators, top memory consumers, and host pressure. Remove unnecessary competing services, tune query memory grants or work areas, adjust engine memory settings, and add RAM or resize the service tier when the working set has outgrown the instance.
If this alert is noisy, check whether the workload naturally has short bursts. Raise the duration or threshold for expected batch windows, but keep a lower route for production systems where user-facing latency matters. Avoid simply disabling the alert until you know whether the noise is caused by threshold choice or by a real capacity trend.