If
health_monitor.memory_running=ON,
Health Monitor samples system memory and the MySQL server process
at the interval specified by
health_monitor.memory_sample_rate.
Measurements are stored in the Performance Schema tables
health_system_memory and
health_process_memory. The default
memory sampling interval is 60 seconds and the default retention
is 86400 seconds, providing capacity for 1440 measurements in each
table.
The system memory table distinguishes free memory from available memory. Available memory is the platform-defined estimate used for warning thresholds. On Linux it estimates memory available for new applications without swapping. Memory consumption by other processes on the host can therefore cause Health Monitor memory warnings. Platform-specific metrics can be NULL when unavailable.
The default memory sampling interval is 60 seconds and the default retention is 86400 seconds, providing capacity for 1440 measurements in each table.
health_monitor.memory_usage_warning_level_1,
health_monitor.memory_usage_warning_level_2
and
health_monitor.memory_usage_warning_level_3
define absolute available-memory warning thresholds, with defaults
of 1024, 500, and 100 MiB, respectively. Health Monitor reports
the most severe applicable warning and reports changes as memory
decreases or recovers. These are available-memory thresholds
rather than percentages of process memory in use.
Memory warnings do not automatically shut down the server or terminate queries in response to low memory.