lmxopcua/src/Server/ZB.MOM.WW.OtOpcUa.AdminUI/Components/Pages/FleetDiagnostics.razor

@page "/fleet/diagnostics"
@* Per-node driver diagnostics, fetched on demand from the selected driver node's DriverHostActor over
   Akka (IFleetDiagnosticsClient → GetDiagnostics → NodeDiagnosticsSnapshot). The node answers from
   its live child set + each child's last-published health; the "running from cache" flag is the one
   fact /hosts cannot show (a node booted from its artifact cache looks healthy but is frozen).
   Reach it as /fleet/diagnostics?node=<host:port> (the NodeId shown on /fleet and /hosts), or pick
   a driver-role member from the selector. *@
@attribute [Authorize(Policy = AdminUiPolicies.AuthenticatedRead)]
@rendermode RenderMode.InteractiveServer
@using Microsoft.EntityFrameworkCore
@using ZB.MOM.WW.OtOpcUa.AdminUI.Clients
@using ZB.MOM.WW.OtOpcUa.Cluster
@using ZB.MOM.WW.OtOpcUa.Commons.Interfaces
@using ZB.MOM.WW.OtOpcUa.Commons.Types
@using ZB.MOM.WW.OtOpcUa.Configuration
@inject IClusterRoleInfo Cluster
@inject IFleetDiagnosticsClient Diagnostics
@inject IDbContextFactory<OtOpcUaConfigDbContext> DbFactory
@inject NavigationManager Nav
@implements IDisposable

<PageErrorBoundary>
<div class="d-flex justify-content-between align-items-center mb-3">
    <h4 class="mb-0">Node diagnostics</h4>
    <a class="btn btn-sm btn-outline-secondary" href="/fleet">Back to fleet status</a>
</div>

<div class="d-flex align-items-center mb-3 gap-2 flex-wrap">
    <label class="small text-muted mb-0" for="diag-node">Node</label>
    <select id="diag-node" class="form-select form-select-sm" style="width:auto"
            value="@_selectedNode" @onchange="OnNodeSelected">
        <option value="">— select a driver node —</option>
        @foreach (var member in _members)
        {
            <option value="@member">@member</option>
        }
        @if (!string.IsNullOrEmpty(_selectedNode) && !_members.Contains(_selectedNode))
        {
            <option value="@_selectedNode">@_selectedNode (not an Up member)</option>
        }
    </select>
    <input type="text" class="form-control form-control-sm" style="width:220px"
           placeholder="or type host:port…" @bind="_typedNode" @bind:event="oninput"
           @onkeydown="OnTypedKeyDown" />
    <button class="btn btn-sm btn-outline-primary" @onclick="UseTypedNodeAsync"
            disabled="@(string.IsNullOrWhiteSpace(_typedNode))">Go</button>
    <button class="btn btn-sm btn-outline-primary" @onclick="RefreshAsync"
            disabled="@(_loading || string.IsNullOrEmpty(_selectedNode))">
        @if (_loading) { <span class="spinner-border spinner-border-sm me-1" /> }
        Refresh
    </button>
    <span class="text-muted small">
        Last fetched: @(_fetchedUtc?.ToString("HH:mm:ss 'UTC'") ?? "—")
    </span>
</div>

@if (string.IsNullOrEmpty(_selectedNode))
{
    <section class="panel notice rise" style="animation-delay:.02s">
        Pick a driver-role node to fetch its driver diagnostics. The node answers from its own
        DriverHostActor, so this works for any node reachable over Akka remoting from this admin node.
    </section>
}
else if (_snapshot is null)
{
    <p>@(_loading ? "Fetching…" : "No snapshot yet.")</p>
}
else
{
    <section class="agg-grid rise" style="animation-delay:.02s">
        <div class="agg-card">
            <div class="agg-label">Node</div>
            <div class="agg-value mono">@_snapshot.NodeId.Value</div>
        </div>
        <div class="agg-card">
            <div class="agg-label">Revision</div>
            <div class="agg-value mono" title="@(_snapshot.CurrentRevision?.Value ?? "")">
                @(_snapshot.CurrentRevision is { } rev ? DisplayText.Abbreviate(rev.Value, 12) : "—")
            </div>
        </div>
        <div class="agg-card @(_snapshot.RunningFromCache ? "caution" : "")">
            <div class="agg-label">Config source</div>
            <div class="agg-value">@(_snapshot.RunningFromCache ? "cache (frozen)" : "live")</div>
        </div>
        <div class="agg-card">
            <div class="agg-label">Drivers</div>
            <div class="agg-value numeric">@_snapshot.Drivers.Count</div>
        </div>
        <div class="agg-card alert">
            <div class="agg-label">Faulted devices</div>
            <div class="agg-value numeric">@_snapshot.Drivers.Sum(d => d.FaultedDevices)</div>
        </div>
    </section>

    @if (_snapshot.RunningFromCache)
    {
        <section class="panel notice rise" style="animation-delay:.05s">
            This node booted its configuration from its local artifact cache because central was unreachable.
            It serves its last-known-good address space, but no deployment can reach it until central is back.
        </section>
    }

    <section class="panel rise" style="animation-delay:.08s">
        <div class="panel-head">
            Drivers
            <span class="text-muted small ms-2">as of @_snapshot.AsOfUtc.ToString("u")</span>
        </div>
        @if (_snapshot.Drivers.Count == 0)
        {
            <div class="p-3 text-muted">
                Empty snapshot: the node has no driver children — or it did not answer within the client's
                3 s budget (an unreachable node degrades to an empty snapshot; see the AdminUI log for the
                FleetDiagnosticsClient warning).
            </div>
        }
        else
        {
            <div class="table-wrap">
                <table class="data-table">
                    <thead>
                        <tr>
                            <th>Driver instance</th>
                            <th>Row id</th>
                            <th>State</th>
                            <th class="numeric">Connected devices</th>
                            <th class="numeric">Faulted devices</th>
                            <th title="When the driver child last published a changed health snapshot (its spawn time until the first)">Last health publish</th>
                        </tr>
                    </thead>
                    <tbody>
                        @foreach (var d in _snapshot.Drivers.OrderBy(d => DisplayNameOf(d), StringComparer.OrdinalIgnoreCase))
                        {
                            <tr>
                                <td>
                                    @* A-F5: the wire carries the string instance id; show the authored name
                                       (joined from the ConfigDb) with the id beneath it. *@
                                    @DisplayNameOf(d)
                                    @if (!string.Equals(DisplayNameOf(d), d.Name, StringComparison.Ordinal))
                                    {
                                        <div class="mono small text-muted">@d.Name</div>
                                    }
                                </td>
                                <td><span class="mono" title="@d.DriverInstanceId">@(d.DriverInstanceId == Guid.Empty ? "—" : DisplayText.Abbreviate(d.DriverInstanceId.ToString(), 13))</span></td>
                                <td><span class="chip @StateChipClass(d.State)">@d.State</span></td>
                                <td class="numeric">@d.ConnectedDevices</td>
                                <td class="numeric">@d.FaultedDevices</td>
                                <td>@d.LastChangeUtc.ToString("u")</td>
                            </tr>
                        }
                    </tbody>
                </table>
            </div>
        }
    </section>
}
</PageErrorBoundary>

@code {
    [SupplyParameterFromQuery(Name = "node")]
    public string? Node { get; set; }

    private IReadOnlyList<string> _members = Array.Empty<string>();
    private string _selectedNode = "";
    private string _typedNode = "";
    private NodeDiagnosticsSnapshot? _snapshot;
    private IReadOnlyDictionary<string, string> _displayNames = new Dictionary<string, string>();
    private bool _loading;
    private DateTime? _fetchedUtc;

    // Page-lifetime cancellation: a fetch in flight when the circuit goes away is abandoned.
    private readonly CancellationTokenSource _cts = new();
    private volatile bool _disposed;

    protected override async Task OnParametersSetAsync()
    {
        _members = Cluster.MembersWithRole(RoleParser.Driver)
            .Select(n => n.Value)
            .OrderBy(s => s, StringComparer.OrdinalIgnoreCase)
            .ToList();
        var requested = Node?.Trim() ?? "";
        if (!string.Equals(requested, _selectedNode, StringComparison.Ordinal))
        {
            _selectedNode = requested;
            _snapshot = null;
            if (!string.IsNullOrEmpty(_selectedNode)) await LoadAsync();
        }
    }

    private Task OnNodeSelected(ChangeEventArgs e)
    {
        var value = e.Value?.ToString() ?? "";
        Navigate(value);
        return Task.CompletedTask;
    }

    private Task UseTypedNodeAsync()
    {
        Navigate(_typedNode.Trim());
        return Task.CompletedTask;
    }

    private Task OnTypedKeyDown(KeyboardEventArgs e)
    {
        if (e.Key == "Enter") Navigate(_typedNode.Trim());
        return Task.CompletedTask;
    }

    // The node id lives in the URL so a diagnostics view is linkable from /fleet and survives a reload.
    private void Navigate(string node) =>
        Nav.NavigateTo(string.IsNullOrEmpty(node)
            ? "/fleet/diagnostics"
            : FleetDiagnosticsLinks.For(node));

    private async Task RefreshAsync() => await LoadAsync();

    private async Task LoadAsync()
    {
        if (_loading || string.IsNullOrEmpty(_selectedNode)) return;
        _loading = true;
        try
        {
            // The client is fail-safe by contract: a node that does not answer yields an empty snapshot.
            _snapshot = await Diagnostics.GetDiagnosticsAsync(NodeId.Parse(_selectedNode), _cts.Token);
            _fetchedUtc = DateTime.UtcNow;
            _displayNames = await LoadDisplayNamesAsync(_snapshot);
        }
        catch (Exception ex) when (ex is OperationCanceledException or ObjectDisposedException)
        {
            // Circuit gone mid-fetch — a clean exit, not a fault.
        }
        finally
        {
            _loading = false;
            if (!_disposed) StateHasChanged();
        }
    }

    // The authored name for a driver row, else its id (no ConfigDb row, or the name lookup failed).
    private string DisplayNameOf(DriverInstanceDiagnostics d) =>
        _displayNames.TryGetValue(d.Name, out var name) ? name : d.Name;

    private async Task<IReadOnlyDictionary<string, string>> LoadDisplayNamesAsync(NodeDiagnosticsSnapshot snapshot)
    {
        try
        {
            await using var db = await DbFactory.CreateDbContextAsync(_cts.Token);
            return await FleetDiagnosticsLinks.LoadDisplayNamesAsync(
                db, snapshot.Drivers.Select(d => d.Name).ToList(), _cts.Token);
        }
        catch (Exception ex) when (ex is not OperationCanceledException and not ObjectDisposedException)
        {
            // Names are enrichment only: a ConfigDb hiccup falls back to the ids, never fails the page.
            return new Dictionary<string, string>();
        }
    }

    private static string StateChipClass(string state) => state switch
    {
        "Healthy" => "chip-ok",
        "Degraded" or "Connecting" or "Reconnecting" => "chip-caution",
        "Faulted" => "chip-alert",
        _ => "chip-idle",
    };

    public void Dispose()
    {
        _disposed = true;
        _cts.Cancel();
        _cts.Dispose();
    }
}