dotfiles/agent/src/system/health.rs

//! System health module
use std::{collections::HashMap, sync::Arc};

use chrono::{DateTime, Utc};
use tokio::sync::watch;

use super::SYSTEM;

const MEMORY_USAGE_CRITICAL_THRESHOLD: u64 = 90;
const CPU_USAGE_CRITICAL_THRESHOLD: u64 = 90;
const DISK_USAGE_CRITICAL_THRESHOLD: u64 = 90;

#[derive(Clone, PartialEq)]
pub enum CriticalReason {
    HighMemoryUsage,
    HighCpuUsage,
    HighDiskUsage,
}

#[derive(Clone, Default, PartialEq)]
pub enum SystemStatus {
    #[default]
    Normal,
    OutOfDate,
    Updating,
    Critical(Vec<CriticalReason>),
}

#[derive(Clone, Default)]
pub struct SystemHealth {
    pub status: SystemStatus,
}

#[derive(Clone, PartialEq, Debug)]
pub enum TaskStatus {
    Normal,
    Warning,
    Error,
    Completed,
}

#[derive(Clone)]
pub struct TaskHealth {
    status: TaskStatus,
    started_on: DateTime<Utc>,
    message: String,
    progress: u8,
}

impl TaskHealth {
    pub fn new(message: String) -> Self {
        let started_on = chrono::Utc::now();

        Self {
            status: TaskStatus::Normal,
            started_on,
            message,
            progress: 0,
        }
    }

    pub fn set_normal(&mut self, message: String) {
        self.status = TaskStatus::Normal;
        self.message = message;
    }

    pub fn set_warning(&mut self, message: String) {
        self.status = TaskStatus::Warning;
        self.message = message;
    }

    pub fn set_error(&mut self, message: String) {
        self.status = TaskStatus::Error;
        self.message = message;
    }

    pub fn set_completed(mut self, message: String) {
        self.status = TaskStatus::Completed;
        self.progress = 100;
        self.message = message;
    }

    pub fn set_progress(&mut self, message: String, progress: u8) {
        self.progress = progress;
        self.message = message;
    }

    pub fn status(&self) -> &TaskStatus {
        &self.status
    }

    pub fn started_on(&self) -> &DateTime<Utc> {
        &self.started_on
    }

    pub fn message(&self) -> &str {
        &self.message
    }

    pub fn progress(&self) -> u8 {
        self.progress
    }
}

#[derive(Default, Clone)]
pub struct Health {
    system: SystemHealth,
    tasks: HashMap<String, TaskHealth>,
}

impl Health {
    pub fn system(&self) -> SystemHealth {
        self.system.clone()
    }

    pub fn tasks(self) -> HashMap<String, TaskHealth> {
        self.tasks
    }
}

/// `HealthMonitor` gives access to shared system health state, allowing to watch health and update
/// task health status.
///
/// # Usage
/// Internally `HealthMonitor` uses [Arc] so it can be cheaply cloned and shared.
///
/// ```no_run
/// use prymn_agent::system::health::{HealthMonitor, TaskHealth};
///
/// let health_monitor = HealthMonitor::new();
/// let health_monitor_clone = health_monitor.clone();
/// tokio::spawn(async move {
///     loop {
///         health_monitor_clone.check_system().await;
///     }
/// });
/// tokio::spawn(async move {
///     health_monitor.set_task_health(
///         "some_task".to_string(),
///         TaskHealth::new("example".to_string())
///     );
/// });
/// ```
#[derive(Clone)]
pub struct HealthMonitor {
    sender: Arc<watch::Sender<Health>>,
    receiver: watch::Receiver<Health>,
}

impl HealthMonitor {
    pub fn new() -> Self {
        let (sender, receiver) = watch::channel(Health::default());
        Self {
            sender: Arc::new(sender),
            receiver,
        }
    }

    // TODO: Remove async from here (so it can be consistent)
    //       Move system checking task into it's own thing
    pub async fn check_system(&self) {
        use sysinfo::{CpuExt, DiskExt, SystemExt};

        let status = tokio::task::spawn_blocking(|| {
            let mut status = SystemStatus::Normal;

            // TODO: For testability, dependency inject this System struct in this function.
            let mut sys = SYSTEM.lock().unwrap();

            // Refresh system resources usage
            sys.refresh_specifics(
                sysinfo::RefreshKind::new()
                    .with_memory()
                    .with_disks()
                    .with_cpu(sysinfo::CpuRefreshKind::new().with_cpu_usage()),
            );

            let mut statuses = vec![];

            // Check for critical memory usage
            let memory_usage = sys.used_memory() * 100 / sys.total_memory();
            if memory_usage > MEMORY_USAGE_CRITICAL_THRESHOLD {
                statuses.push(CriticalReason::HighMemoryUsage);
            }

            // Check for critical CPU usage
            let cpu_usage = sys.global_cpu_info().cpu_usage();
            if cpu_usage > CPU_USAGE_CRITICAL_THRESHOLD as f32 {
                statuses.push(CriticalReason::HighCpuUsage);
            }

            // Check for any disk usage that is critical
            for disk in sys.disks() {
                let available_disk = disk.available_space() * 100 / disk.total_space();
                if available_disk < 100 - DISK_USAGE_CRITICAL_THRESHOLD {
                    statuses.push(CriticalReason::HighDiskUsage);
                }
            }

            if !statuses.is_empty() {
                status = SystemStatus::Critical(statuses);
            }

            status
        })
        .await
        .expect("system checking task panicked - possibly due to panicked mutex lock");

        self.sender.send_if_modified(|Health { system, .. }| {
            if system.status == status {
                return false;
            }

            system.status = status;
            true
        });
    }

    pub fn set_task_health(&self, task_name: String, health: TaskHealth) {
        // Always send a notification in this case since it is an explicit action.
        self.sender.send_modify(|Health { tasks, .. }| {
            tasks.insert(task_name, health);
        });
    }

    pub fn clear_task(&self, task_name: &str) {
        self.sender
            .send_if_modified(|Health { tasks, .. }| tasks.remove(task_name).is_some());
    }

    pub fn monitor(&self) -> watch::Receiver<Health> {
        self.receiver.clone()
    }
}

impl Default for HealthMonitor {
    fn default() -> Self {
        HealthMonitor::new()
    }
}

impl std::fmt::Display for SystemStatus {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            SystemStatus::Normal => write!(f, "normal"),
            SystemStatus::OutOfDate => write!(f, "out of date"),
            SystemStatus::Updating => write!(f, "updating"),
            SystemStatus::Critical(_) => write!(f, "critical"),
        }
    }
}

impl std::fmt::Display for CriticalReason {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            CriticalReason::HighMemoryUsage => write!(f, "high memory usage"),
            CriticalReason::HighCpuUsage => write!(f, "high cpu usage"),
            CriticalReason::HighDiskUsage => write!(f, "high disk usage"),
        }
    }
}

impl std::fmt::Display for TaskStatus {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            TaskStatus::Normal => write!(f, "normal"),
            TaskStatus::Warning => write!(f, "warning"),
            TaskStatus::Error => write!(f, "error"),
            TaskStatus::Completed => write!(f, "completed"),
        }
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_task_monitor() {
        let health_monitor = HealthMonitor::new();
        let receiver = health_monitor.monitor();

        assert!(receiver.has_changed().is_ok_and(|changed| !changed));

        let health = TaskHealth::new("this is normal".to_owned());
        health_monitor.set_task_health("some_task".to_string(), health);

        assert!(receiver.has_changed().is_ok_and(|changed| changed));

        {
            let health = receiver.borrow();
            let task_health = health.tasks.get("some_task").expect("a task should exist");

            assert_eq!(task_health.status, TaskStatus::Normal);
            assert_eq!(task_health.progress, 0);
            assert_eq!(task_health.message, "this is normal");
        }

        health_monitor.clear_task("some_task");
        assert!(!receiver.borrow().tasks.contains_key("some_task"));
    }
}