pub struct AlertsConfig {
pub filesystem_warning_percent: f64,
pub filesystem_critical_percent: f64,
pub filesystem_full_warning_hours: f64,
pub cpu_warning_celsius: f32,
pub gpu_warning_celsius: f32,
pub drive_warning_celsius: f32,
pub memory_warning_percent: f64,
pub gpu_memory_warning_percent: f64,
pub ignored_units: Vec<String>,
}Expand description
Thresholds that turn readings into alerts (overview cards, status dots and notifications). 0 turns an alert off. Applied by the front-end, so remote hosts are judged by the viewer’s settings.
Fields§
§filesystem_warning_percent: f64Filesystem usage (%) for a warning.
filesystem_critical_percent: f64Filesystem usage (%) for a critical alert.
filesystem_full_warning_hours: f64A filesystem that will be full within this many hours, at the growth of the last half hour, is a warning.
cpu_warning_celsius: f32Hottest CPU reading (°C) for a warning. Desktop Ryzen CPUs run up to 95 °C under full load by design.
gpu_warning_celsius: f32GPU temperature (°C) for a warning.
drive_warning_celsius: f32Drive temperature (°C, from SMART or the NVMe sensor) for a warning.
memory_warning_percent: f64Memory in use (%, page cache not counted) for a warning.
gpu_memory_warning_percent: f64GPU memory in use (%) for a warning. Off by default: PyTorch, JAX and TensorFlow reserve most of the VRAM up front, so a busy GPU always looks full.
ignored_units: Vec<String>systemd units whose failure is not worth a warning (e.g. a
systemd-networkd-wait-online.service that fails on every boot).