feat(runtime,adminui): real GetDiagnostics from child health + /fleet/diagnostics consumer (A6)
DriverHostActor wraps the health publisher it hands its children in a ChildHealthLedger (last snapshot per driver id, forwarded to the real publisher), so GetDiagnostics now answers with each child's row id, authored name, last-published state, device counts (host statuses when the driver probes hosts, else one device) and last-change time — instead of Guid.Empty / "Spawned" / 0 / now. Entries are forgotten on StopChild and operator Restart. IFleetDiagnosticsClient gains its first consumer: the new /fleet/diagnostics page (node selector + ?node= query, revision, running-from-cache notice, per-driver table). FleetDiagnosticsClient logs the degrade-to-empty path at Warning. Review ref: 02/U1. Refs #731 Claude-Session: https://claude.ai/code/session_01Ub9hGmYLQvn9XETzw7CMsT
This commit is contained in:
Joseph Doherty
@@ -0,0 +1,238 @@
|
||||
@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 ZB.MOM.WW.OtOpcUa.Cluster
|
||||
@using ZB.MOM.WW.OtOpcUa.Commons.Interfaces
|
||||
@using ZB.MOM.WW.OtOpcUa.Commons.Types
|
||||
@inject IClusterRoleInfo Cluster
|
||||
@inject IFleetDiagnosticsClient Diagnostics
|
||||
@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</th>
|
||||
<th>Row id</th>
|
||||
<th>State</th>
|
||||
<th class="numeric">Connected devices</th>
|
||||
<th class="numeric">Faulted devices</th>
|
||||
<th>Last change</th>
|
||||
</tr>
|
||||
</thead>
|
||||
<tbody>
|
||||
@foreach (var d in _snapshot.Drivers.OrderBy(d => d.Name, StringComparer.OrdinalIgnoreCase))
|
||||
{
|
||||
<tr>
|
||||
<td>@d.Name</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 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"
|
||||
: "/fleet/diagnostics?node=" + Uri.EscapeDataString(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;
|
||||
}
|
||||
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();
|
||||
}
|
||||
}
|
||||
|
||||
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();
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user