Issue #15: implement dashboard snapshot service

2026-04-26 17:49:41 -04:00
parent a20517f5ad
commit 3661420f0a
12 changed files with 602 additions and 0 deletions
@@ -0,0 +1,290 @@
+using Microsoft.Extensions.Options;
+using MxGateway.Contracts.Proto;
+using MxGateway.Server.Configuration;
+using MxGateway.Server.Dashboard;
+using MxGateway.Server.Metrics;
+using MxGateway.Server.Sessions;
+using MxGateway.Server.Workers;
+
+namespace MxGateway.Tests.Gateway.Dashboard;
+
+public sealed class DashboardSnapshotServiceTests
+{
+    [Fact]
+    public void GetSnapshot_WhenRegistryEmpty_ReturnsEmptyOperationalState()
+    {
+        using GatewayMetrics metrics = new();
+        DashboardSnapshotService service = CreateService(new SessionRegistry(), metrics);
+
+        DashboardSnapshot snapshot = service.GetSnapshot();
+
+        Assert.Empty(snapshot.Sessions);
+        Assert.Empty(snapshot.Workers);
+        Assert.Empty(snapshot.Faults);
+        Assert.Contains(snapshot.Metrics, metric => metric.Name == "mxgateway.sessions.open" && metric.Value == 0);
+        Assert.Equal("Healthy", snapshot.GatewayStatus);
+        Assert.NotNull(snapshot.Configuration);
+    }
+
+    [Fact]
+    public void GetSnapshot_ProjectsActiveAndFaultedSessionsWorkersMetricsAndFaults()
+    {
+        SessionRegistry registry = new();
+        GatewaySession activeSession = CreateSession(
+            "session-active",
+            "client-one",
+            DateTimeOffset.Parse("2026-04-26T10:00:00Z"));
+        activeSession.AttachWorkerClient(new FakeWorkerClient("session-active", 1201, WorkerClientState.Ready));
+        activeSession.MarkReady();
+        GatewaySession faultedSession = CreateSession(
+            "session-faulted",
+            "client-two",
+            DateTimeOffset.Parse("2026-04-26T10:01:00Z"));
+        faultedSession.AttachWorkerClient(new FakeWorkerClient("session-faulted", 1202, WorkerClientState.Faulted));
+        faultedSession.MarkFaulted("worker pipe disconnected");
+        registry.TryAdd(activeSession);
+        registry.TryAdd(faultedSession);
+        using GatewayMetrics metrics = new();
+        metrics.SessionOpened();
+        metrics.SessionOpened();
+        metrics.CommandStarted("Register");
+        metrics.CommandFailed("Register", "WorkerFaulted", TimeSpan.FromMilliseconds(7));
+        metrics.EventReceived("session-active", "OnDataChange");
+        metrics.Fault("WorkerFaulted");
+        DashboardSnapshotService service = CreateService(registry, metrics);
+
+        DashboardSnapshot snapshot = service.GetSnapshot();
+
+        Assert.Equal(2, snapshot.Sessions.Count);
+        Assert.Equal("session-faulted", snapshot.Sessions[0].SessionId);
+        Assert.Equal(SessionState.Faulted, snapshot.Sessions[0].State);
+        Assert.Equal(2, snapshot.Workers.Count);
+        Assert.Contains(snapshot.Metrics, metric => metric.Name == "mxgateway.commands.started" && metric.Value == 1);
+        Assert.Contains(
+            snapshot.Metrics,
+            metric => metric.Name == "mxgateway.events.received"
+                && metric.Dimension == "OnDataChange"
+                && metric.Value == 1);
+        DashboardFaultSummary fault = Assert.Single(snapshot.Faults);
+        Assert.Equal("Worker", fault.Source);
+        Assert.Equal("session-faulted", fault.SessionId);
+        Assert.Equal("worker pipe disconnected", fault.Message);
+    }
+
+    [Fact]
+    public void GetSnapshot_RedactsSecretsFromSessionAndFaultFields()
+    {
+        SessionRegistry registry = new();
+        GatewaySession session = CreateSession(
+            "session-redacted",
+            "Bearer mxgw_admin_super-secret",
+            DateTimeOffset.Parse("2026-04-26T10:00:00Z"),
+            clientSessionName: "password=hunter2",
+            clientCorrelationId: "token=abc123");
+        session.MarkFaulted("secret=credential-value");
+        registry.TryAdd(session);
+        using GatewayMetrics metrics = new();
+        DashboardSnapshotService service = CreateService(registry, metrics);
+
+        DashboardSnapshot snapshot = service.GetSnapshot();
+
+        DashboardSessionSummary summary = Assert.Single(snapshot.Sessions);
+        Assert.Equal("Bearer mxgw_admin_[redacted]", summary.ClientIdentity);
+        Assert.Equal("[redacted]", summary.ClientSessionName);
+        Assert.Equal("[redacted]", summary.ClientCorrelationId);
+        Assert.Equal("[redacted]", summary.LastFault);
+        Assert.Equal("[redacted]", Assert.Single(snapshot.Faults).Message);
+        Assert.Equal("[redacted]", snapshot.Configuration.Authentication.PepperSecretName);
+    }
+
+    [Fact]
+    public void GetSnapshot_DoesNotMutateSessionOrWorkerState()
+    {
+        SessionRegistry registry = new();
+        GatewaySession session = CreateSession(
+            "session-active",
+            "client-one",
+            DateTimeOffset.Parse("2026-04-26T10:00:00Z"));
+        FakeWorkerClient workerClient = new("session-active", 1201, WorkerClientState.Ready);
+        session.AttachWorkerClient(workerClient);
+        session.MarkReady();
+        registry.TryAdd(session);
+        using GatewayMetrics metrics = new();
+        DashboardSnapshotService service = CreateService(registry, metrics);
+
+        service.GetSnapshot();
+        service.GetSnapshot();
+
+        Assert.Equal(1, registry.ActiveCount);
+        Assert.Equal(SessionState.Ready, session.State);
+        Assert.Equal(WorkerClientState.Ready, workerClient.State);
+        Assert.Equal(0, workerClient.StartCount);
+        Assert.Equal(0, workerClient.ShutdownCount);
+        Assert.Equal(0, workerClient.KillCount);
+    }
+
+    [Fact]
+    public void GetSnapshot_AppliesRecentSessionAndFaultLimits()
+    {
+        SessionRegistry registry = new();
+        GatewaySession olderSession = CreateSession(
+            "session-older",
+            "client-one",
+            DateTimeOffset.Parse("2026-04-26T10:00:00Z"));
+        GatewaySession newerSession = CreateSession(
+            "session-newer",
+            "client-two",
+            DateTimeOffset.Parse("2026-04-26T10:01:00Z"));
+        olderSession.MarkFaulted("older fault");
+        newerSession.MarkFaulted("newer fault");
+        registry.TryAdd(olderSession);
+        registry.TryAdd(newerSession);
+        using GatewayMetrics metrics = new();
+        DashboardSnapshotService service = CreateService(
+            registry,
+            metrics,
+            new GatewayOptions
+            {
+                Dashboard = new DashboardOptions
+                {
+                    SnapshotIntervalMilliseconds = 1,
+                    RecentSessionLimit = 1,
+                    RecentFaultLimit = 1,
+                },
+            });
+
+        DashboardSnapshot snapshot = service.GetSnapshot();
+
+        Assert.Equal("session-newer", Assert.Single(snapshot.Sessions).SessionId);
+        Assert.Equal("session-newer", Assert.Single(snapshot.Faults).SessionId);
+    }
+
+    [Fact]
+    public async Task WatchSnapshotsAsync_WhenSubscriberCancels_DisposesCleanly()
+    {
+        using GatewayMetrics metrics = new();
+        DashboardSnapshotService service = CreateService(
+            new SessionRegistry(),
+            metrics,
+            new GatewayOptions
+            {
+                Dashboard = new DashboardOptions
+                {
+                    SnapshotIntervalMilliseconds = 1,
+                },
+            });
+        using CancellationTokenSource cancellation = new();
+        await using IAsyncEnumerator<DashboardSnapshot> enumerator = service
+            .WatchSnapshotsAsync(cancellation.Token)
+            .GetAsyncEnumerator();
+
+        Assert.True(await enumerator.MoveNextAsync().AsTask().WaitAsync(TimeSpan.FromSeconds(1)));
+        await cancellation.CancelAsync();
+        bool hasNext = await enumerator.MoveNextAsync().AsTask().WaitAsync(TimeSpan.FromSeconds(1));
+
+        Assert.False(hasNext);
+    }
+
+    private static DashboardSnapshotService CreateService(
+        SessionRegistry registry,
+        GatewayMetrics metrics,
+        GatewayOptions? options = null)
+    {
+        GatewayOptions resolvedOptions = options ?? new GatewayOptions
+        {
+            Dashboard = new DashboardOptions
+            {
+                SnapshotIntervalMilliseconds = 1,
+            },
+        };
+        GatewayConfigurationProvider configurationProvider = new(Options.Create(resolvedOptions));
+
+        return new DashboardSnapshotService(
+            registry,
+            metrics,
+            configurationProvider,
+            Options.Create(resolvedOptions));
+    }
+
+    private static GatewaySession CreateSession(
+        string sessionId,
+        string? clientIdentity,
+        DateTimeOffset openedAt,
+        string? clientSessionName = "test-session",
+        string? clientCorrelationId = "client-correlation")
+    {
+        return new GatewaySession(
+            sessionId,
+            "mxaccess",
+            $"mxaccess-gateway-1-{sessionId}",
+            "nonce",
+            clientIdentity,
+            clientSessionName,
+            clientCorrelationId,
+            TimeSpan.FromSeconds(30),
+            TimeSpan.FromSeconds(5),
+            TimeSpan.FromSeconds(5),
+            openedAt);
+    }
+
+    private sealed class FakeWorkerClient(
+        string sessionId,
+        int? processId,
+        WorkerClientState state) : IWorkerClient
+    {
+        public string SessionId { get; } = sessionId;
+
+        public int? ProcessId { get; } = processId;
+
+        public WorkerClientState State { get; private set; } = state;
+
+        public DateTimeOffset LastHeartbeatAt { get; } = DateTimeOffset.Parse("2026-04-26T10:02:00Z");
+
+        public int StartCount { get; private set; }
+
+        public int ShutdownCount { get; private set; }
+
+        public int KillCount { get; private set; }
+
+        public Task StartAsync(CancellationToken cancellationToken)
+        {
+            StartCount++;
+            return Task.CompletedTask;
+        }
+
+        public Task<WorkerCommandReply> InvokeAsync(
+            WorkerCommand command,
+            TimeSpan timeout,
+            CancellationToken cancellationToken)
+        {
+            return Task.FromResult(new WorkerCommandReply());
+        }
+
+        public async IAsyncEnumerable<WorkerEvent> ReadEventsAsync(
+            [System.Runtime.CompilerServices.EnumeratorCancellation] CancellationToken cancellationToken)
+        {
+            await Task.CompletedTask;
+            yield break;
+        }
+
+        public Task ShutdownAsync(
+            TimeSpan timeout,
+            CancellationToken cancellationToken)
+        {
+            ShutdownCount++;
+            State = WorkerClientState.Closed;
+            return Task.CompletedTask;
+        }
+
+        public void Kill(string reason)
+        {
+            KillCount++;
+            State = WorkerClientState.Faulted;
+        }
+
+        public ValueTask DisposeAsync()
+        {
+            return ValueTask.CompletedTask;
+        }
+    }
+}