rmqtt 0.23.0

//! MQTT Broker Node Management Core
//!
//! Provides centralized node monitoring and resource management for MQTT broker clusters, implementing:
//! 1. **Node State Tracking**:
//!    - Uptime calculation with chrono integration
//!    - System load monitoring (1/5/15-minute averages)
//!    - Memory/Disk usage statistics collection
//! 2. **Cluster Health Management**:
//!    - Busy state detection with configurable thresholds
//!    - CPU load aggregation using systemstat
//!    - Graceful degradation through max_busy_loadavg/max_busy_cpuloadavg
//! 3. **Protocol Implementation**:
//!    - gRPC server/client integration for cluster communication
//!    - JSON serialization of broker/node status (BrokerInfo/NodeInfo)
//!    - Version metadata exposure (Rustc + build version)
//!
//! Key components align with MQTT specification requirements:
//! - Persistent session management through NodeStatus tracking
//! - Resource monitoring for connection capacity planning
//! - Distributed architecture support via gRPC
//!

use std::sync::atomic::{AtomicBool, AtomicI64, Ordering};
use std::time::Duration;

use serde::{Deserialize, Serialize};
use serde_json::json;
use systemstat::Platform;

#[cfg(feature = "grpc")]
use crate::context::CircuitBreakerConfig;
use crate::context::ServerContext;
#[cfg(feature = "grpc")]
use crate::grpc::{GrpcClient, GrpcServer};
use crate::types::{NodeId, TimestampMillis};
use crate::utils::timestamp_millis;

const VERSION: &str = env!("CARGO_PKG_VERSION");

const RUSTC_VERSION: &str = env!("RUSTC_VERSION");

const RUSTC_BUILD_TIME: &str = env!("RUSTC_BUILD_TIME");

/// Represents a single RMQTT broker node.
///
/// Tracks identity, runtime metadata, and system resource utilization
/// for health monitoring and cluster coordination.
///
/// # Busy Detection
///
/// A node is considered "busy" when either:
/// - System load average (1-min) exceeds `max_busy_loadavg`
/// - CPU load exceeds `max_busy_cpuloadavg`
///
/// Busy state is cached with a configurable update interval to
/// avoid excessive system calls on the hot path.
#[derive(Debug)]
pub struct Node {
    pub id: NodeId,
    pub start_time: chrono::DateTime<chrono::Local>,
    max_busy_loadavg: f32,
    max_busy_cpuloadavg: f32,
    pub(crate) busy_update_interval: Duration,
    cpuload: AtomicI64,
}

impl Default for Node {
    fn default() -> Self {
        Self::new(0, 80.0, 90.0, Duration::from_secs(2))
    }
}

impl Node {
    pub fn new(
        id: NodeId,
        max_busy_loadavg: f32,
        max_busy_cpuloadavg: f32,
        busy_update_interval: Duration,
    ) -> Self {
        Self {
            id,
            start_time: chrono::Local::now(),
            max_busy_loadavg,
            max_busy_cpuloadavg,
            busy_update_interval,
            cpuload: AtomicI64::new(0),
        }
    }

    #[inline]
    pub fn id(&self) -> NodeId {
        self.id
    }

    #[inline]
    pub async fn name(&self, scx: &ServerContext, id: NodeId) -> String {
        scx.extends.shared().await.node_name(id)
    }

    #[cfg(feature = "grpc")]
    #[inline]
    pub async fn new_grpc_client(
        &self,
        remote_addr: &str,
        connect_timeout: Duration,
        client_concurrency_limit: usize,
        _batch_size: usize,
        cb_config: &CircuitBreakerConfig,
    ) -> crate::Result<GrpcClient> {
        GrpcClient::new(remote_addr, connect_timeout, client_concurrency_limit, cb_config).await
    }

    #[cfg(feature = "grpc")]
    pub fn start_grpc_server(
        &self,
        scx: ServerContext,
        server_addr: std::net::SocketAddr,
        reuseaddr: bool,
        reuseport: bool,
    ) {
        tokio::spawn(async move {
            if let Err(e) = GrpcServer::new(scx).listen_and_serve(server_addr, reuseaddr, reuseport).await {
                log::error!("listen and serve failure, {e}, laddr: {server_addr:?}");
            }
        });
    }

    #[inline]
    pub async fn status(&self, scx: &ServerContext) -> NodeStatus {
        match scx.extends.shared().await.health_status().await {
            Ok(status) => {
                if status.running {
                    NodeStatus::Running(1)
                } else {
                    NodeStatus::Stop
                }
            }
            Err(e) => NodeStatus::Error(e.to_string()),
        }
    }

    #[inline]
    fn uptime(&self) -> String {
        to_uptime((chrono::Local::now() - self.start_time).num_seconds())
    }

    #[inline]
    pub fn version() -> String {
        format!("rmqtt/{VERSION}-{RUSTC_BUILD_TIME}")
    }

    #[inline]
    pub fn rustc_version() -> String {
        RUSTC_VERSION.to_string()
    }

    #[inline]
    pub async fn broker_info(&self, scx: &ServerContext) -> BrokerInfo {
        let node_id = self.id;
        BrokerInfo {
            version: Self::version(),
            rustc_version: Self::rustc_version(),
            uptime: self.uptime(),
            sysdescr: "RMQTT Broker".into(),
            node_status: self.status(scx).await,
            node_id,
            node_name: self.name(scx, node_id).await, //Runtime::instance().extends.shared().await.node_name(node_id),
            datetime: chrono::Local::now().format("%Y-%m-%d %H:%M:%S").to_string(),
        }
    }

    #[inline]
    pub async fn node_info(&self, scx: &ServerContext) -> NodeInfo {
        let node_id = self.id;

        let sys = systemstat::System::new();
        let boottime = sys.boot_time().map(|t| t.to_string()).unwrap_or_default();
        let loadavg = sys.load_average();
        let mem_info = sys.memory();

        let (disk_total, disk_free) = if let Ok(mounts) = sys.mounts() {
            let total = mounts.iter().map(|m| m.total.as_u64()).sum();
            let free = mounts.iter().map(|m| m.free.as_u64()).sum();
            (total, free)
        } else {
            (0, 0)
        };

        NodeInfo {
            connections: scx.connections.count(),
            boottime,
            load1: loadavg.as_ref().map(|l| l.one).unwrap_or_default(),
            load5: loadavg.as_ref().map(|l| l.five).unwrap_or_default(),
            load15: loadavg.as_ref().map(|l| l.fifteen).unwrap_or_default(),
            memory_total: mem_info.as_ref().map(|m| m.total.as_u64()).unwrap_or_default(),
            memory_free: mem_info.as_ref().map(|m| m.free.as_u64()).unwrap_or_default(),
            memory_used: mem_info
                .as_ref()
                .map(|m| systemstat::saturating_sub_bytes(m.total, m.free).as_u64())
                .unwrap_or_default(),
            disk_total,
            disk_free,
            node_status: self.status(scx).await,
            node_id,
            node_name: self.name(scx, node_id).await, //Runtime::instance().extends.shared().await.node_name(node_id),
            uptime: self.uptime(),
            version: Self::version(),
            rustc_version: Self::rustc_version(),
        }
    }

    #[inline]
    fn _is_busy(&self) -> bool {
        let sys = systemstat::System::new();
        let cpuload = self.cpuload();

        let loadavg = sys.load_average();
        let load1 = loadavg.as_ref().map(|l| l.one).unwrap_or_default();

        load1 > self.max_busy_loadavg || cpuload > self.max_busy_cpuloadavg
    }

    #[inline]
    pub fn sys_is_busy(&self) -> bool {
        static CACHED_BUSY: AtomicBool = AtomicBool::new(false);
        static CACHED_TIME: AtomicI64 = AtomicI64::new(0);

        let now = timestamp_millis();

        let last_update = CACHED_TIME.load(Ordering::Relaxed);

        if now - last_update < self.busy_update_interval.as_millis() as TimestampMillis {
            return CACHED_BUSY.load(Ordering::Relaxed);
        }

        let busy = self._is_busy();

        CACHED_BUSY.store(busy, Ordering::Relaxed);
        CACHED_TIME.store(now, Ordering::Relaxed);

        busy
    }

    #[inline]
    pub fn cpuload(&self) -> f32 {
        //0.0 - 100.0
        self.cpuload.load(Ordering::SeqCst) as f32 / 100.0
    }

    pub async fn update_cpuload(&self) {
        let sys = systemstat::System::new();
        let cpuload_aggr = sys.cpu_load_aggregate().ok();
        tokio::time::sleep(Duration::from_secs(2)).await;
        let cpuload_aggr = cpuload_aggr.and_then(|dm| dm.done().ok());
        let cpuload = cpuload_aggr
            .map(|cpuload_aggr| {
                let aggregate1 =
                    cpuload_aggr.user + cpuload_aggr.nice + cpuload_aggr.system + cpuload_aggr.interrupt;
                aggregate1 * 100.0
            })
            .unwrap_or_default();
        self.cpuload.store(cpuload as i64, Ordering::SeqCst);
    }
}

/// Broker-level metadata exposed via the management API.
///
/// Includes version information, uptime, node identity, and
/// current operational status.
#[derive(Serialize, Deserialize, Clone, Debug)]
pub struct BrokerInfo {
    pub version: String,
    pub rustc_version: String,
    pub uptime: String,
    pub sysdescr: String,
    pub node_status: NodeStatus,
    pub node_id: NodeId,
    pub node_name: String,
    pub datetime: String,
}

impl BrokerInfo {
    pub fn to_json(&self) -> serde_json::Value {
        json!({
            "version": self.version,
            "rustc_version": self.rustc_version,
            "uptime": self.uptime,
            "sysdescr": self.sysdescr,
            "running": self.node_status.is_running(),
            "node_id": self.node_id,
            "node_name": self.node_name,
            "datetime": self.datetime
        })
    }
}

/// Detailed per-node system resource information.
///
/// Captures CPU load averages, memory and disk usage, connection
/// counts, and runtime metadata for cluster monitoring.
#[derive(Serialize, Deserialize, Clone, Debug, Default)]
pub struct NodeInfo {
    pub connections: isize,
    pub boottime: String,
    pub load1: f32,
    pub load5: f32,
    pub load15: f32,
    // pub max_fds: usize,
    // pub cpu_num: String,
    // pub cpu_speed: String,
    pub memory_total: u64,
    pub memory_used: u64,
    pub memory_free: u64,
    pub disk_total: u64,
    pub disk_free: u64,
    // pub os_release: String,
    // pub os_type: String,
    // pub proc_total: String,
    pub node_status: NodeStatus,
    pub node_id: NodeId,
    pub node_name: String,
    pub uptime: String,
    pub version: String,
    pub rustc_version: String,
}

impl NodeInfo {
    #[inline]
    pub fn to_json(&self) -> serde_json::Value {
        json!({
            "connections":  self.connections,
            "boottime":  self.boottime,
            "load1":  self.load1,
            "load5":  self.load5,
            "load15":  self.load15,
            // "max_fds":  self.max_fds,
            // "cpu_num":  self.cpu_num,
            // "cpu_speed":  self.cpu_speed,
            "memory_total":  self.memory_total,
            "memory_used":  self.memory_used,
            "memory_free":  self.memory_free,
            "disk_total":  self.disk_total,
            "disk_free":  self.disk_free,
            // "os_release":  self.os_release,
            // "os_type":  self.os_type,
            // "proc_total":  self.proc_total,
            "running":  self.node_status.is_running(),
            "node_id":  self.node_id,
            "node_name":  self.node_name,
            "uptime":  self.uptime,
            "version":  self.version,
            "rustc_version": self.rustc_version,
        })
    }

    #[inline]
    pub fn add(&mut self, other: &NodeInfo) {
        self.load1 += other.load1;
        self.load5 += other.load5;
        self.load15 += other.load15;
        self.memory_total += other.memory_total;
        self.memory_used += other.memory_used;
        self.memory_free += other.memory_free;
        self.disk_total += other.disk_total;
        self.disk_free += other.disk_free;
        self.node_status = {
            let c = match (&self.node_status, &other.node_status) {
                (NodeStatus::Running(c1), NodeStatus::Running(c2)) => *c1 + *c2,
                (NodeStatus::Running(c1), _) => *c1,
                (_, NodeStatus::Running(c2)) => *c2,
                (_, _) => 0,
            };
            NodeStatus::Running(c)
        };
    }
}

/// Operational status of a broker node.
///
/// - `Running(usize)` — Active with an associated count (e.g., connected clients).
/// - `Stop` — Node has been stopped.
/// - `Error(String)` — Node encountered a fault condition.
#[derive(Serialize, Deserialize, Clone, Debug)]
#[serde(rename_all = "lowercase")]
pub enum NodeStatus {
    Running(usize),
    Stop,
    Error(String),
}

impl NodeStatus {
    #[inline]
    pub fn running(&self) -> usize {
        if let NodeStatus::Running(c) = self {
            *c
        } else {
            0
        }
    }

    #[inline]
    pub fn is_running(&self) -> bool {
        matches!(self, NodeStatus::Running(_))
    }
}

impl Default for NodeStatus {
    #[inline]
    fn default() -> Self {
        NodeStatus::Stop
    }
}

/// Convert a duration in seconds to a human-readable uptime string.
///
/// Returns a formatted string like `"2 days 5 hours, 30 minutes, 15 seconds"`.
#[inline]
pub fn to_uptime(uptime: i64) -> String {
    let uptime_secs = uptime % 60;
    let uptime = uptime / 60;
    let uptime_minus = uptime % 60;
    let uptime = uptime / 60;
    let uptime_hours = uptime % 24;
    let uptime_days = uptime / 24;
    format!("{uptime_days} days {uptime_hours} hours, {uptime_minus} minutes, {uptime_secs} seconds")
}