lmxopcua/src/Server/ZB.MOM.WW.OtOpcUa.Admin/Components/Pages/Clusters/RedundancyTab.razor

@using Microsoft.AspNetCore.SignalR.Client
@using ZB.MOM.WW.OtOpcUa.Admin.Hubs
@using ZB.MOM.WW.OtOpcUa.Admin.Services
@using ZB.MOM.WW.OtOpcUa.Configuration.Entities
@using ZB.MOM.WW.OtOpcUa.Configuration.Enums
@inject ClusterNodeService NodeSvc
@inject NavigationManager Nav
@inject AdminHubConnectionFactory HubFactory
@implements IAsyncDisposable

<h4 class="panel-head">Redundancy topology</h4>
@if (_roleChangedBanner is not null)
{
    <section class="panel notice rise" style="animation-delay:.02s">@_roleChangedBanner</section>
}
<p class="text-muted small">
    One row per <span class="mono">ClusterNode</span> in this cluster. Role, <span class="mono">ApplicationUri</span>,
    and <span class="mono">ServiceLevelBase</span> are authored separately; the Admin UI shows them read-only
    here so operators can confirm the published topology without touching it. LastSeen older than
    @((int)ClusterNodeService.StaleThreshold.TotalSeconds)s is flagged Stale — the node has
    stopped heart-beating and is likely down. Role swap goes through the server-side
    <span class="mono">RedundancyCoordinator</span> apply-lease flow, not direct DB edits.
</p>

@if (_nodes is null)
{
    <p>Loading…</p>
}
else if (_nodes.Count == 0)
{
    <section class="panel notice rise" style="animation-delay:.02s">
        No ClusterNode rows for this cluster. The server process needs at least one entry
        (with a non-blank <span class="mono">ApplicationUri</span>) before it can start up per OPC UA spec.
    </section>
}
else
{
    var primaries = _nodes.Count(n => n.RedundancyRole == RedundancyRole.Primary);
    var secondaries = _nodes.Count(n => n.RedundancyRole == RedundancyRole.Secondary);
    var standalone = _nodes.Count(n => n.RedundancyRole == RedundancyRole.Standalone);
    var staleCount = _nodes.Count(ClusterNodeService.IsStale);

    <section class="agg-grid rise" style="animation-delay:.02s">
        <div class="agg-card">
            <div class="agg-label">Nodes</div>
            <div class="agg-value numeric">@_nodes.Count</div>
        </div>
        <div class="agg-card">
            <div class="agg-label">Primary</div>
            <div class="agg-value numeric @(primaries > 0 ? "s-ok" : "")">@primaries</div>
        </div>
        <div class="agg-card">
            <div class="agg-label">Secondary</div>
            <div class="agg-value numeric">@secondaries</div>
        </div>
        <div class="agg-card">
            <div class="agg-label">Stale</div>
            <div class="agg-value numeric @(staleCount > 0 ? "s-warn" : "")">@staleCount</div>
        </div>
    </section>

    @if (primaries == 0 && standalone == 0)
    {
        <section class="panel notice rise" style="animation-delay:.08s">
            <span class="s-bad">No Primary or Standalone node — the cluster has no authoritative write target. Secondaries
            stay read-only until one of them gets promoted via <span class="mono">RedundancyCoordinator</span>.</span>
        </section>
    }
    else if (primaries > 1)
    {
        <section class="panel notice rise" style="animation-delay:.08s">
            <span class="s-bad"><strong>Split-brain:</strong> @primaries nodes claim the Primary role. Apply-lease
            enforcement should have made this impossible at the coordinator level. Investigate
            immediately — one of the rows was likely hand-edited.</span>
        </section>
    }

    <section class="panel rise" style="animation-delay:.14s">
        <div class="table-wrap">
            <table class="data-table">
                <thead>
                    <tr>
                        <th>Node</th>
                        <th>Role</th>
                        <th>Host</th>
                        <th class="num">OPC UA port</th>
                        <th class="num">ServiceLevel base</th>
                        <th>ApplicationUri</th>
                        <th>Enabled</th>
                        <th>Last seen</th>
                    </tr>
                </thead>
                <tbody>
                @foreach (var n in _nodes)
                {
                    <tr>
                        <td><span class="mono">@n.NodeId</span></td>
                        <td><span class="chip @RoleBadge(n.RedundancyRole)">@n.RedundancyRole</span></td>
                        <td>@n.Host</td>
                        <td class="num mono">@n.OpcUaPort</td>
                        <td class="num">@n.ServiceLevelBase</td>
                        <td class="mono">@n.ApplicationUri</td>
                        <td>
                            @if (n.Enabled) { <span class="chip chip-ok">Enabled</span> }
                            else { <span class="chip chip-idle">Disabled</span> }
                        </td>
                        <td class="@(ClusterNodeService.IsStale(n) ? "s-warn" : "")">
                            @(n.LastSeenAt is null ? "never" : FormatAge(n.LastSeenAt.Value))
                            @if (ClusterNodeService.IsStale(n)) { <span class="chip chip-warn ms-1">Stale</span> }
                        </td>
                    </tr>
                }
                </tbody>
            </table>
        </div>
    </section>
}

@code {
    [Parameter] public string ClusterId { get; set; } = string.Empty;

    private List<ClusterNode>? _nodes;
    private HubConnection? _hub;
    private string? _roleChangedBanner;

    protected override async Task OnParametersSetAsync()
    {
        _nodes = await NodeSvc.ListByClusterAsync(ClusterId, CancellationToken.None);
        if (_hub is null) await ConnectHubAsync();
    }

    private async Task ConnectHubAsync()
    {
        _hub = HubFactory.Create("/hubs/fleet");

        _hub.On<RoleChangedMessage>("RoleChanged", async msg =>
        {
            if (msg.ClusterId != ClusterId) return;
            _roleChangedBanner = $"Role changed on {msg.NodeId}: {msg.FromRole} → {msg.ToRole} at {msg.ObservedAtUtc:HH:mm:ss 'UTC'}";
            _nodes = await NodeSvc.ListByClusterAsync(ClusterId, CancellationToken.None);
            await InvokeAsync(StateHasChanged);
        });

        // Best-effort: FleetStatusHub requires an authenticated caller, and the server-side
        // HubConnection cannot forward the browser auth cookie — swallow connect failures so
        // the tab still renders. Live role-change updates degrade.
        try
        {
            await _hub.StartAsync();
            await _hub.SendAsync("SubscribeCluster", ClusterId);
        }
        catch
        {
            // best-effort live updates — see comment above
        }
    }

    public async ValueTask DisposeAsync()
    {
        if (_hub is not null)
        {
            await _hub.DisposeAsync();
            _hub = null;
        }
    }

    private static string RowClass(ClusterNode n) =>
        ClusterNodeService.IsStale(n) ? "table-warning" :
        !n.Enabled ? "table-secondary" : "";

    private static string RoleBadge(RedundancyRole r) => r switch
    {
        RedundancyRole.Primary    => "chip-ok",
        RedundancyRole.Secondary  => "chip-idle",
        RedundancyRole.Standalone => "chip-idle",
        _                         => "chip-idle",
    };

    private static string FormatAge(DateTime t)
    {
        var age = DateTime.UtcNow - t;
        if (age.TotalSeconds < 60) return $"{(int)age.TotalSeconds}s ago";
        if (age.TotalMinutes < 60) return $"{(int)age.TotalMinutes}m ago";
        if (age.TotalHours < 24) return $"{(int)age.TotalHours}h ago";
        return t.ToString("yyyy-MM-dd HH:mm 'UTC'");
    }
}