Migu2.0/MiGu.Server/Fleet/FleetHealthService.cs

using System.Text.Json;
using Microsoft.Extensions.Options;
using MiGu.Server.Auth;
using MiGu.Server.Launcher;
using MiGu.Server.Ota;

namespace MiGu.Server.Fleet;

/// <summary>
/// 车队健康：保留 SimpleLite 的 CPU/故障率等，延迟改为对 WatchDog(:9776) 做 TCP RTT。
/// SimpleLite /fleet/health 探测的是车载 HTTP :8081，多数现场未开该端口会假超时 2000ms。
/// </summary>
public sealed class FleetHealthService
{
    private readonly IHttpClientFactory _httpFactory;
    private readonly SimpleLiteOptions _sl;
    private readonly InternalTokenStore _token;
    private readonly WatchDogClient _wd;
    private readonly ILogger<FleetHealthService> _log;

    private static readonly JsonSerializerOptions JsonOpt = new()
    {
        PropertyNameCaseInsensitive = true
    };

    public FleetHealthService(
        IHttpClientFactory httpFactory,
        IOptions<SimpleLiteOptions> sl,
        InternalTokenStore token,
        WatchDogClient wd,
        ILogger<FleetHealthService> log)
    {
        _httpFactory = httpFactory;
        _sl = sl.Value;
        _token = token;
        _wd = wd;
        _log = log;
    }

    public async Task<List<FleetHealthRowDto>> GetAsync(CancellationToken ct)
    {
        var rows = await FetchSimpleLiteHealthAsync(ct);
        if (rows.Count == 0)
            rows = await BuildRowsFromCarsAsync(ct);

        await Parallel.ForEachAsync(
            rows,
            new ParallelOptions { MaxDegreeOfParallelism = 8, CancellationToken = ct },
            async (row, token) =>
            {
                if (string.IsNullOrWhiteSpace(row.Ip))
                {
                    // 无 IP 时保留 SimpleLite 原探测结果
                    row.LatencySource ??= row.LatencyMs != null ? "onboard" : "none";
                    return;
                }

                // 取两次 TCP 连接的较小值，降低偶发握手抖动
                var a = await _wd.MeasureRttMsAsync(row.Ip, token);
                var b = await _wd.MeasureRttMsAsync(row.Ip, token);
                int? rtt = (a, b) switch
                {
                    (null, null) => null,
                    (int x, null) => x,
                    (null, int y) => y,
                    (int x, int y) => Math.Min(x, y)
                };

                if (rtt != null)
                {
                    row.LatencyMs = rtt;
                    row.Reachable = true;
                    row.LatencySource = "watchdog";
                    row.ProbedAt = DateTimeOffset.UtcNow.ToString("O");
                }
                else
                {
                    // WatchDog 不通：若 SimpleLite 车载探测成功则保留，否则标不可达
                    if (row.Reachable == true && row.LatencyMs is > 0 and < 2000)
                    {
                        row.LatencySource = "onboard";
                    }
                    else
                    {
                        row.Reachable = false;
                        row.LatencyMs = null;
                        row.LatencySource = "watchdog";
                        row.ProbedAt = DateTimeOffset.UtcNow.ToString("O");
                    }
                }
            });

        return rows;
    }

    private async Task<List<FleetHealthRowDto>> FetchSimpleLiteHealthAsync(CancellationToken ct)
    {
        var port = _sl.ProjectionPort > 0 ? _sl.ProjectionPort : 8222;
        try
        {
            using var client = _httpFactory.CreateClient();
            client.Timeout = TimeSpan.FromSeconds(20);
            using var req = new HttpRequestMessage(HttpMethod.Get, $"http://127.0.0.1:{port}/projection/fleet/health");
            var token = _token.Token;
            if (!string.IsNullOrEmpty(token))
                req.Headers.TryAddWithoutValidation("X-Platform-Internal-Token", token);

            using var resp = await client.SendAsync(req, ct);
            if (!resp.IsSuccessStatusCode) return new();

            var text = await resp.Content.ReadAsStringAsync(ct);
            var list = JsonSerializer.Deserialize<List<FleetHealthRowDto>>(text, JsonOpt);
            return list ?? new();
        }
        catch (Exception ex)
        {
            _log.LogDebug(ex, "fleet/health from SimpleLite failed");
            return new();
        }
    }

    private async Task<List<FleetHealthRowDto>> BuildRowsFromCarsAsync(CancellationToken ct)
    {
        var port = _sl.ProjectionPort > 0 ? _sl.ProjectionPort : 8222;
        try
        {
            using var client = _httpFactory.CreateClient();
            client.Timeout = TimeSpan.FromSeconds(10);
            using var req = new HttpRequestMessage(HttpMethod.Get, $"http://127.0.0.1:{port}/projection/cars");
            var token = _token.Token;
            if (!string.IsNullOrEmpty(token))
                req.Headers.TryAddWithoutValidation("X-Platform-Internal-Token", token);

            using var resp = await client.SendAsync(req, ct);
            if (!resp.IsSuccessStatusCode) return new();

            var text = await resp.Content.ReadAsStringAsync(ct);
            using var doc = JsonDocument.Parse(text);
            if (doc.RootElement.ValueKind != JsonValueKind.Array) return new();

            var list = new List<FleetHealthRowDto>();
            foreach (var el in doc.RootElement.EnumerateArray())
            {
                var rawId = el.TryGetProperty("rawId", out var rid) && rid.TryGetInt32(out var id)
                    ? id
                    : 0;
                if (rawId <= 0) continue;
                var ip = el.TryGetProperty("ip", out var ipEl) ? ipEl.GetString() : null;
                var name = el.TryGetProperty("name", out var nEl) ? nEl.GetString() : null;
                var onboard = el.TryGetProperty("onboardUrl", out var oEl) ? oEl.GetString() : null;
                if (string.IsNullOrEmpty(onboard) && !string.IsNullOrEmpty(ip))
                    onboard = $"http://{ip}:8081";
                list.Add(new FleetHealthRowDto
                {
                    CarId = rawId,
                    CarName = name,
                    Ip = ip,
                    OnboardUrl = onboard
                });
            }
            return list;
        }
        catch (Exception ex)
        {
            _log.LogDebug(ex, "projection/cars fallback for fleet health failed");
            return new();
        }
    }
}