natsdotnet/docs/plans/2026-03-12-e2e-cluster-tests-plan.md

# E2E Cluster & RAFT Tests Implementation Plan

> **For Claude:** REQUIRED SUB-SKILL: Use superpowers-extended-cc:executing-plans to implement this plan task-by-task.

**Goal:** Create `NATS.E2E.Cluster.Tests` with 16 E2E tests covering cluster resilience, RAFT consensus, JetStream cluster replication, gateway failover, and leaf node failover.

**Architecture:** Single xUnit 3 project, 4 fixtures (ThreeNodeCluster, JetStreamCluster, GatewayPair, HubLeaf), all running real multi-process servers via `NatsServerProcess`.

**Tech Stack:** .NET 10, xUnit 3, Shouldly, NATS.Client.Core, NATS.Client.JetStream, NATS.NKeys

---

### Task 0: Create project scaffold and add to solution

**Files:**
- Create: `tests/NATS.E2E.Cluster.Tests/NATS.E2E.Cluster.Tests.csproj`
- Modify: `NatsDotNet.slnx`

**Step 1: Create the csproj file**

```xml
<Project Sdk="Microsoft.NET.Sdk">

  <PropertyGroup>
    <IsPackable>false</IsPackable>
  </PropertyGroup>

  <ItemGroup>
    <PackageReference Include="coverlet.collector" />
    <PackageReference Include="Microsoft.NET.Test.Sdk" />
    <PackageReference Include="NATS.Client.Core" />
    <PackageReference Include="NATS.Client.JetStream" />
    <PackageReference Include="NATS.NKeys" />
    <PackageReference Include="Shouldly" />
    <PackageReference Include="xunit" />
    <PackageReference Include="xunit.runner.visualstudio" />
  </ItemGroup>

  <ItemGroup>
    <Using Include="Xunit" />
    <Using Include="Shouldly" />
  </ItemGroup>

</Project>
```

**Step 2: Add to solution file**

Add `<Project Path="tests/NATS.E2E.Cluster.Tests/NATS.E2E.Cluster.Tests.csproj" />` to the `/tests/` folder in `NatsDotNet.slnx`.

**Step 3: Verify build**

Run: `dotnet build tests/NATS.E2E.Cluster.Tests`
Expected: Build succeeded, 0 errors

**Step 4: Commit**

```bash
git add tests/NATS.E2E.Cluster.Tests/NATS.E2E.Cluster.Tests.csproj NatsDotNet.slnx
git commit -m "feat: scaffold NATS.E2E.Cluster.Tests project"
```

---

### Task 1: Copy NatsServerProcess and create base infrastructure

**Files:**
- Create: `tests/NATS.E2E.Cluster.Tests/Infrastructure/NatsServerProcess.cs`

**Step 1: Copy NatsServerProcess from NATS.E2E.Tests**

Copy `tests/NATS.E2E.Tests/Infrastructure/NatsServerProcess.cs` to `tests/NATS.E2E.Cluster.Tests/Infrastructure/NatsServerProcess.cs`. Change the namespace from `NATS.E2E.Tests.Infrastructure` to `NATS.E2E.Cluster.Tests.Infrastructure`.

**Step 2: Verify build**

Run: `dotnet build tests/NATS.E2E.Cluster.Tests`
Expected: Build succeeded

**Step 3: Commit**

```bash
git add tests/NATS.E2E.Cluster.Tests/Infrastructure/NatsServerProcess.cs
git commit -m "feat: add NatsServerProcess to cluster E2E infrastructure"
```

---

### Task 2: Create ThreeNodeClusterFixture with KillNode/RestartNode

**Files:**
- Create: `tests/NATS.E2E.Cluster.Tests/Infrastructure/ThreeNodeClusterFixture.cs`

**Step 1: Write the fixture**

Model after existing `ClusterFixture` from `NATS.E2E.Tests` but with these additions:

- Store `NatsServerProcess[]` array (3 servers) and parallel `string[]` for config content and `int[]` for cluster ports
- Pre-allocate all ports (client, cluster, monitoring) in constructor or `InitializeAsync`
- Full-mesh routes config with `pool_size: 1`, monitoring enabled, `-DV` flags
- `WaitForFullMeshAsync()` — polls `/routez` on all alive nodes until each reports `NumRoutes >= expectedRouteCount`
- `KillNode(int index)` — calls `DisposeAsync()` on that server process, nulls it out
- `RestartNode(int index)` — creates a new `NatsServerProcess` using the saved config content and ports, calls `StartAsync()`
- `WaitForFullMeshAsync(int expectedRoutes = 2)` — parameterized so after a kill, we can wait for `NumRoutes >= 1` on survivors
- `CreateClient(int index)` — returns `NatsConnection` to specified node
- `GetServerOutput(int index)` — returns output from specified node
- Private JSON model `Routez` with `num_routes` property

The key difference from the existing `ClusterFixture`: the fixture must support restarting servers on the same ports. This means:
- The `NatsServerProcess` constructor takes a pre-assigned port. We need a new constructor or factory method. **However**, looking at the existing `NatsServerProcess`, it allocates its own port in the constructor. For the kill/restart pattern, we need the port to be stable. The simplest approach: add a new `NatsServerProcess` constructor overload that accepts a pre-assigned port. Copy this into our local version.

Add to `NatsServerProcess`:
```csharp
/// <summary>
/// Constructor with pre-assigned port for kill/restart scenarios.
/// </summary>
public NatsServerProcess(int port, string[]? extraArgs = null, string? configContent = null, bool enableMonitoring = false, int? monitorPort = null)
{
    Port = port;
    _extraArgs = extraArgs;
    _configContent = configContent;
    _enableMonitoring = enableMonitoring;
    if (monitorPort.HasValue)
        MonitorPort = monitorPort.Value;
    else if (_enableMonitoring)
        MonitorPort = AllocateFreePort();
}
```

Fixture pattern:
```csharp
namespace NATS.E2E.Cluster.Tests.Infrastructure;

public sealed class ThreeNodeClusterFixture : IAsyncLifetime
{
    private NatsServerProcess?[] _servers = new NatsServerProcess?[3];
    private readonly int[] _clientPorts = new int[3];
    private readonly int[] _clusterPorts = new int[3];
    private readonly int[] _monitorPorts = new int[3];
    private readonly string[] _configs = new string[3];

    public int GetPort(int index) => _clientPorts[index];
    public int GetMonitorPort(int index) => _monitorPorts[index];

    public async Task InitializeAsync() { /* allocate ports, build configs, start 3, wait mesh */ }
    public async Task KillNode(int index) { /* dispose server, null it */ }
    public async Task RestartNode(int index) { /* recreate server on same ports, start */ }
    public async Task WaitForFullMeshAsync(int expectedRoutes = 2) { /* poll /routez */ }
    public NatsConnection CreateClient(int index) { /* ... */ }
    public string GetServerOutput(int index) { /* ... */ }
    public async Task DisposeAsync() { /* dispose all */ }
}
```

**Step 2: Write a smoke test to verify the fixture**

Create a temporary test (can be removed later or kept):
```csharp
// In a temp file or ClusterResilienceTests.cs
[Fact]
public async Task Fixture_Smoke_ThreeNodeClusterStarts()
{
    // Just verifies the fixture starts without errors — implicit via IAsyncLifetime
    await Task.CompletedTask; // fixture.InitializeAsync already ran
}
```

**Step 3: Run the smoke test**

Run: `dotnet test tests/NATS.E2E.Cluster.Tests --filter "FullyQualifiedName~Fixture_Smoke" -v normal`
Expected: PASS

**Step 4: Commit**

```bash
git add tests/NATS.E2E.Cluster.Tests/Infrastructure/
git commit -m "feat: add ThreeNodeClusterFixture with KillNode/RestartNode"
```

---

### Task 3: Create JetStreamClusterFixture

**Files:**
- Create: `tests/NATS.E2E.Cluster.Tests/Infrastructure/JetStreamClusterFixture.cs`

**Step 1: Write the fixture**

Same as `ThreeNodeClusterFixture` but with `jetstream { store_dir, max_mem_store, max_file_store }` in each node's config. Store dirs are created in `/tmp/nats-e2e-js-cluster-{guid}-node{i}` and deleted on dispose.

Key differences from `ThreeNodeClusterFixture`:
- Config includes `jetstream { store_dir: "{storeDir}" max_mem_store: 64mb max_file_store: 256mb }`
- `DisposeAsync` also deletes store directories
- `RestartNode` does NOT delete the store dir (preserves state for catchup tests)

Pattern:
```csharp
public sealed class JetStreamClusterFixture : IAsyncLifetime
{
    // Same fields as ThreeNodeClusterFixture plus:
    private readonly string[] _storeDirs = new string[3];

    // Config template adds jetstream block
    // Dispose cleans up store dirs
    // RestartNode preserves store dirs
}
```

**Step 2: Verify build**

Run: `dotnet build tests/NATS.E2E.Cluster.Tests`
Expected: Build succeeded

**Step 3: Commit**

```bash
git add tests/NATS.E2E.Cluster.Tests/Infrastructure/JetStreamClusterFixture.cs
git commit -m "feat: add JetStreamClusterFixture for R3 replication tests"
```

---

### Task 4: Create GatewayPairFixture with KillNode/RestartNode

**Files:**
- Create: `tests/NATS.E2E.Cluster.Tests/Infrastructure/GatewayPairFixture.cs`

**Step 1: Write the fixture**

Model after existing `GatewayFixture` from `NATS.E2E.Tests` but with:
- Pre-allocated ports (client, gateway, monitoring) for both servers
- `KillNode(int index)` / `RestartNode(int index)` (index 0 = server A, 1 = server B)
- `WaitForGatewayConnectionAsync()` — polls `/gatewayz` on alive servers until `NumGateways >= 1`
- `CreateClientA()` / `CreateClientB()`

```csharp
public sealed class GatewayPairFixture : IAsyncLifetime
{
    private NatsServerProcess?[] _servers = new NatsServerProcess?[2];
    private readonly int[] _clientPorts = new int[2];
    private readonly int[] _gwPorts = new int[2];
    private readonly int[] _monitorPorts = new int[2];
    private readonly string[] _configs = new string[2];

    public async Task KillNode(int index) { /* ... */ }
    public async Task RestartNode(int index) { /* ... */ }
    public async Task WaitForGatewayConnectionAsync() { /* ... */ }
    // ...
}
```

**Step 2: Verify build**

Run: `dotnet build tests/NATS.E2E.Cluster.Tests`
Expected: Build succeeded

**Step 3: Commit**

```bash
git add tests/NATS.E2E.Cluster.Tests/Infrastructure/GatewayPairFixture.cs
git commit -m "feat: add GatewayPairFixture for failover tests"
```

---

### Task 5: Create HubLeafFixture with KillNode/RestartNode

**Files:**
- Create: `tests/NATS.E2E.Cluster.Tests/Infrastructure/HubLeafFixture.cs`

**Step 1: Write the fixture**

Model after existing `LeafNodeFixture` from `NATS.E2E.Tests` but with:
- Pre-allocated ports (client, leaf-listen, monitoring) for hub and leaf
- `KillNode(int index)` / `RestartNode(int index)` (index 0 = hub, 1 = leaf)
- `WaitForLeafConnectionAsync()` — polls hub's `/leafz` until `NumLeafs >= 1`
- Hub starts first in `InitializeAsync()`, leaf starts after hub is ready
- `RestartNode(0)` for hub: restart hub, then wait for leaf to reconnect
- `RestartNode(1)` for leaf: restart leaf, then wait for hub to detect connection

```csharp
public sealed class HubLeafFixture : IAsyncLifetime
{
    private NatsServerProcess?[] _servers = new NatsServerProcess?[2]; // 0=hub, 1=leaf
    private readonly int[] _clientPorts = new int[2];
    private readonly int _leafListenPort;
    private readonly int _hubMonitorPort;
    private readonly string[] _configs = new string[2];

    public int HubPort => _clientPorts[0];
    public int LeafPort => _clientPorts[1];

    public async Task KillNode(int index) { /* ... */ }
    public async Task RestartNode(int index) { /* ... */ }
    public async Task WaitForLeafConnectionAsync() { /* ... */ }
    public NatsConnection CreateHubClient() { /* ... */ }
    public NatsConnection CreateLeafClient() { /* ... */ }
}
```

**Step 2: Verify build**

Run: `dotnet build tests/NATS.E2E.Cluster.Tests`
Expected: Build succeeded

**Step 3: Commit**

```bash
git add tests/NATS.E2E.Cluster.Tests/Infrastructure/HubLeafFixture.cs
git commit -m "feat: add HubLeafFixture for leaf node failover tests"
```

---

### Task 6: Write ClusterResilienceTests (4 tests)

**Files:**
- Create: `tests/NATS.E2E.Cluster.Tests/ClusterResilienceTests.cs`

**Step 1: Write the failing tests**

```csharp
using NATS.Client.Core;
using NATS.E2E.Cluster.Tests.Infrastructure;

namespace NATS.E2E.Cluster.Tests;

public class ClusterResilienceTests(ThreeNodeClusterFixture fixture) : IClassFixture<ThreeNodeClusterFixture>
{
    [Fact]
    public async Task NodeDies_TrafficReroutesToSurvivors()
    {
        // Kill node 2
        await fixture.KillNode(2);

        // Pub on node 0, sub on node 1 — should still work
        await using var pub = fixture.CreateClient(0);
        await using var sub = fixture.CreateClient(1);
        await pub.ConnectAsync();
        await sub.ConnectAsync();

        await using var subscription = await sub.SubscribeCoreAsync<string>("e2e.resilience.reroute");
        await sub.PingAsync();

        // Wait for route between node 0 and node 1 to stabilize
        await fixture.WaitForFullMeshAsync(expectedRoutes: 1);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(10));
        await WaitForPropagationAsync(pub, sub, "e2e.resilience.reroute.probe", cts.Token);

        await pub.PublishAsync("e2e.resilience.reroute", "still-alive", cancellationToken: cts.Token);
        var msg = await subscription.Msgs.ReadAsync(cts.Token);
        msg.Data.ShouldBe("still-alive");

        // Restart node 2 for fixture cleanup
        await fixture.RestartNode(2);
        await fixture.WaitForFullMeshAsync();
    }

    [Fact]
    public async Task NodeRejoins_SubscriptionsPropagateAgain()
    {
        // Kill node 2
        await fixture.KillNode(2);

        // Restart node 2
        await fixture.RestartNode(2);
        await fixture.WaitForFullMeshAsync();

        // Subscribe on restarted node 2, publish on node 0
        await using var pub = fixture.CreateClient(0);
        await using var sub = fixture.CreateClient(2);
        await pub.ConnectAsync();
        await sub.ConnectAsync();

        await using var subscription = await sub.SubscribeCoreAsync<string>("e2e.resilience.rejoin");
        await sub.PingAsync();

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(10));
        await WaitForPropagationAsync(pub, sub, "e2e.resilience.rejoin.probe", cts.Token);

        await pub.PublishAsync("e2e.resilience.rejoin", "welcome-back", cancellationToken: cts.Token);
        var msg = await subscription.Msgs.ReadAsync(cts.Token);
        msg.Data.ShouldBe("welcome-back");
    }

    [Fact]
    public async Task AllRoutesReconnect_AfterNodeRestart()
    {
        await fixture.KillNode(1);
        await fixture.RestartNode(1);
        await fixture.WaitForFullMeshAsync(); // All 3 nodes report NumRoutes >= 2

        // Verify messaging works across all nodes
        await using var pub = fixture.CreateClient(0);
        await using var sub = fixture.CreateClient(1);
        await pub.ConnectAsync();
        await sub.ConnectAsync();

        await using var subscription = await sub.SubscribeCoreAsync<string>("e2e.resilience.reconnect");
        await sub.PingAsync();

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(10));
        await WaitForPropagationAsync(pub, sub, "e2e.resilience.reconnect.probe", cts.Token);

        await pub.PublishAsync("e2e.resilience.reconnect", "reconnected", cancellationToken: cts.Token);
        var msg = await subscription.Msgs.ReadAsync(cts.Token);
        msg.Data.ShouldBe("reconnected");
    }

    [Fact]
    public async Task QueueGroup_NodeDies_RemainingMembersDeliver()
    {
        // Both queue subs on node 1 (same node), publisher on node 0
        await using var pub = fixture.CreateClient(0);
        await using var sub1 = fixture.CreateClient(1);
        await using var sub2 = fixture.CreateClient(1);
        await pub.ConnectAsync();
        await sub1.ConnectAsync();
        await sub2.ConnectAsync();

        await using var s1 = await sub1.SubscribeCoreAsync<int>("e2e.resilience.qg", queueGroup: "rq");
        await using var s2 = await sub2.SubscribeCoreAsync<int>("e2e.resilience.qg", queueGroup: "rq");
        await sub1.PingAsync();
        await sub2.PingAsync();

        using var propCts = new CancellationTokenSource(TimeSpan.FromSeconds(10));
        await WaitForPropagationAsync(pub, sub1, "e2e.resilience.qg.probe", propCts.Token);

        // Kill node 2 (doesn't affect subs on node 1, but tests cluster stability)
        await fixture.KillNode(2);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(10));
        var count = 0;
        var allReceived = new TaskCompletionSource(TaskCreationOptions.RunContinuationsAsynchronously);

        async Task Collect(INatsSub<int> sub, CancellationToken ct)
        {
            await foreach (var _ in sub.Msgs.ReadAllAsync(ct))
            {
                if (Interlocked.Increment(ref count) >= 10)
                    allReceived.TrySetResult();
            }
        }

        _ = Collect(s1, cts.Token);
        _ = Collect(s2, cts.Token);

        for (var i = 0; i < 10; i++)
            await pub.PublishAsync("e2e.resilience.qg", i, cancellationToken: cts.Token);
        await pub.PingAsync(cts.Token);

        await allReceived.Task.WaitAsync(cts.Token);
        count.ShouldBe(10);

        // Restore node 2
        await fixture.RestartNode(2);
        await fixture.WaitForFullMeshAsync();
    }

    /// <summary>
    /// Probe-based propagation wait — same pattern as NATS.E2E.Tests.ClusterTests.
    /// </summary>
    private static async Task WaitForPropagationAsync(
        NatsConnection publisher, NatsConnection subscriber, string probeSubject, CancellationToken ct)
    {
        await using var probeSub = await subscriber.SubscribeCoreAsync<string>(probeSubject, cancellationToken: ct);
        await subscriber.PingAsync(ct);

        using var timer = new PeriodicTimer(TimeSpan.FromMilliseconds(150));
        while (await timer.WaitForNextTickAsync(ct))
        {
            await publisher.PublishAsync(probeSubject, "probe", cancellationToken: ct);
            await publisher.PingAsync(ct);

            using var readCts = CancellationTokenSource.CreateLinkedTokenSource(ct);
            readCts.CancelAfter(TimeSpan.FromMilliseconds(500));
            try
            {
                await probeSub.Msgs.ReadAsync(readCts.Token);
                return; // Propagated
            }
            catch (OperationCanceledException) when (!ct.IsCancellationRequested)
            {
                // Not yet propagated — retry
            }
        }
    }
}
```

**Step 2: Run tests**

Run: `dotnet test tests/NATS.E2E.Cluster.Tests --filter "FullyQualifiedName~ClusterResilience" -v normal`
Expected: All 4 PASS

**Step 3: Commit**

```bash
git add tests/NATS.E2E.Cluster.Tests/ClusterResilienceTests.cs
git commit -m "test: add cluster resilience E2E tests (node failure, rejoin, reconnect, queue groups)"
```

---

### Task 7: Write RaftConsensusTests (4 tests)

**Files:**
- Create: `tests/NATS.E2E.Cluster.Tests/RaftConsensusTests.cs`

**Step 1: Write the tests**

These tests use `JetStreamClusterFixture`. RAFT runs underneath JetStream — creating an R3 stream triggers RAFT group formation. We verify leader election, log replication, and leader failover by querying stream info via `$JS.API`.

```csharp
using NATS.Client.Core;
using NATS.Client.JetStream;
using NATS.Client.JetStream.Models;
using NATS.E2E.Cluster.Tests.Infrastructure;

namespace NATS.E2E.Cluster.Tests;

public class RaftConsensusTests(JetStreamClusterFixture fixture) : IClassFixture<JetStreamClusterFixture>
{
    [Fact]
    public async Task LeaderElection_ClusterFormsLeader()
    {
        // Create R3 stream on node 0
        await using var client = fixture.CreateClient(0);
        await client.ConnectAsync();
        var js = new NatsJSContext(client);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(30));

        var stream = await js.CreateStreamAsync(
            new StreamConfig("RAFT_LEADER", ["raft.leader.>"]) { NumReplicas = 3 },
            cts.Token);

        // Verify stream was created and has cluster info
        stream.Info.Config.Name.ShouldBe("RAFT_LEADER");
        // At least one node should report as having the stream
        stream.Info.State.ShouldNotBeNull();
    }

    [Fact]
    public async Task LeaderDies_NewLeaderElected()
    {
        await using var client0 = fixture.CreateClient(0);
        await client0.ConnectAsync();
        var js0 = new NatsJSContext(client0);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(30));

        await js0.CreateStreamAsync(
            new StreamConfig("RAFT_FAILOVER", ["raft.failover.>"]) { NumReplicas = 3 },
            cts.Token);

        // Publish some data
        for (int i = 0; i < 5; i++)
            await js0.PublishAsync("raft.failover.data", $"msg-{i}", cancellationToken: cts.Token);

        // Kill node 0 (may or may not be leader)
        await fixture.KillNode(0);

        // Try accessing stream from node 1
        await using var client1 = fixture.CreateClient(1);
        await client1.ConnectAsync();
        var js1 = new NatsJSContext(client1);

        // Give RAFT time to elect new leader
        await Task.Delay(2000, cts.Token);

        // Stream should still be accessible — RAFT elected a new leader
        var info = await js1.GetStreamAsync("RAFT_FAILOVER", cancellationToken: cts.Token);
        info.Info.State.Messages.ShouldBeGreaterThanOrEqualTo(5);

        // Restore
        await fixture.RestartNode(0);
        await fixture.WaitForFullMeshAsync();
    }

    [Fact]
    public async Task LogReplication_AllReplicasHaveData()
    {
        await using var client = fixture.CreateClient(0);
        await client.ConnectAsync();
        var js = new NatsJSContext(client);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(30));

        await js.CreateStreamAsync(
            new StreamConfig("RAFT_REPL", ["raft.repl.>"]) { NumReplicas = 3 },
            cts.Token);

        // Publish 10 messages
        for (int i = 0; i < 10; i++)
            await js.PublishAsync("raft.repl.data", $"msg-{i}", cancellationToken: cts.Token);

        // Give replication time to propagate
        await Task.Delay(1000, cts.Token);

        // Query stream info from each node — all should report 10 messages
        for (int node = 0; node < 3; node++)
        {
            await using var c = fixture.CreateClient(node);
            await c.ConnectAsync();
            var nodeJs = new NatsJSContext(c);
            var info = await nodeJs.GetStreamAsync("RAFT_REPL", cancellationToken: cts.Token);
            info.Info.State.Messages.ShouldBe(10, $"Node {node} should have 10 messages");
        }
    }

    [Fact]
    public async Task LeaderRestart_RejoinsAsFollower()
    {
        await using var client0 = fixture.CreateClient(0);
        await client0.ConnectAsync();
        var js0 = new NatsJSContext(client0);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(30));

        await js0.CreateStreamAsync(
            new StreamConfig("RAFT_REJOIN", ["raft.rejoin.>"]) { NumReplicas = 3 },
            cts.Token);

        for (int i = 0; i < 5; i++)
            await js0.PublishAsync("raft.rejoin.data", $"msg-{i}", cancellationToken: cts.Token);

        // Kill node 0
        await fixture.KillNode(0);

        // Publish more on node 1 while node 0 is down
        await using var client1 = fixture.CreateClient(1);
        await client1.ConnectAsync();
        var js1 = new NatsJSContext(client1);

        await Task.Delay(2000, cts.Token); // Wait for RAFT re-election

        for (int i = 5; i < 10; i++)
            await js1.PublishAsync("raft.rejoin.data", $"msg-{i}", cancellationToken: cts.Token);

        // Restart node 0 — it should rejoin and catch up
        await fixture.RestartNode(0);
        await fixture.WaitForFullMeshAsync();

        // Give time for RAFT catchup
        await Task.Delay(2000, cts.Token);

        // Verify node 0 has all 10 messages
        await using var client0b = fixture.CreateClient(0);
        await client0b.ConnectAsync();
        var js0b = new NatsJSContext(client0b);
        var info = await js0b.GetStreamAsync("RAFT_REJOIN", cancellationToken: cts.Token);
        info.Info.State.Messages.ShouldBe(10);
    }
}
```

**Step 2: Run tests**

Run: `dotnet test tests/NATS.E2E.Cluster.Tests --filter "FullyQualifiedName~RaftConsensus" -v normal`
Expected: All 4 PASS

**Step 3: Commit**

```bash
git add tests/NATS.E2E.Cluster.Tests/RaftConsensusTests.cs
git commit -m "test: add RAFT consensus E2E tests (leader election, failover, replication, rejoin)"
```

---

### Task 8: Write JetStreamClusterTests (4 tests)

**Files:**
- Create: `tests/NATS.E2E.Cluster.Tests/JetStreamClusterTests.cs`

**Step 1: Write the tests**

```csharp
using NATS.Client.Core;
using NATS.Client.JetStream;
using NATS.Client.JetStream.Models;
using NATS.E2E.Cluster.Tests.Infrastructure;

namespace NATS.E2E.Cluster.Tests;

public class JetStreamClusterTests(JetStreamClusterFixture fixture) : IClassFixture<JetStreamClusterFixture>
{
    [Fact]
    public async Task R3Stream_CreateAndPublish_ReplicatedAcrossNodes()
    {
        await using var client = fixture.CreateClient(0);
        await client.ConnectAsync();
        var js = new NatsJSContext(client);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(30));

        await js.CreateStreamAsync(
            new StreamConfig("JS_REPL", ["js.repl.>"]) { NumReplicas = 3 },
            cts.Token);

        for (int i = 0; i < 5; i++)
            await js.PublishAsync("js.repl.data", $"msg-{i}", cancellationToken: cts.Token);

        await Task.Delay(1000, cts.Token);

        // All nodes should report 5 messages
        for (int node = 0; node < 3; node++)
        {
            await using var c = fixture.CreateClient(node);
            await c.ConnectAsync();
            var nodeJs = new NatsJSContext(c);
            var info = await nodeJs.GetStreamAsync("JS_REPL", cancellationToken: cts.Token);
            info.Info.State.Messages.ShouldBe(5, $"Node {node} should have 5 messages");
        }
    }

    [Fact]
    public async Task R3Stream_NodeDies_PublishContinues()
    {
        await using var client = fixture.CreateClient(0);
        await client.ConnectAsync();
        var js = new NatsJSContext(client);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(30));

        await js.CreateStreamAsync(
            new StreamConfig("JS_NODEDIES", ["js.nodedies.>"]) { NumReplicas = 3 },
            cts.Token);

        for (int i = 0; i < 3; i++)
            await js.PublishAsync("js.nodedies.data", $"msg-{i}", cancellationToken: cts.Token);

        // Kill node 2
        await fixture.KillNode(2);
        await Task.Delay(2000, cts.Token);

        // Continue publishing on node 0 — should still work with 2/3 nodes
        for (int i = 3; i < 8; i++)
            await js.PublishAsync("js.nodedies.data", $"msg-{i}", cancellationToken: cts.Token);

        // Verify on surviving node 1
        await using var client1 = fixture.CreateClient(1);
        await client1.ConnectAsync();
        var js1 = new NatsJSContext(client1);
        await Task.Delay(1000, cts.Token);
        var info = await js1.GetStreamAsync("JS_NODEDIES", cancellationToken: cts.Token);
        info.Info.State.Messages.ShouldBeGreaterThanOrEqualTo(8);

        await fixture.RestartNode(2);
        await fixture.WaitForFullMeshAsync();
    }

    [Fact]
    public async Task Consumer_NodeDies_PullContinuesOnSurvivor()
    {
        await using var client = fixture.CreateClient(0);
        await client.ConnectAsync();
        var js = new NatsJSContext(client);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(30));

        var stream = await js.CreateStreamAsync(
            new StreamConfig("JS_CONS_FAIL", ["js.consfail.>"]) { NumReplicas = 3 },
            cts.Token);

        for (int i = 0; i < 5; i++)
            await js.PublishAsync("js.consfail.data", $"msg-{i}", cancellationToken: cts.Token);

        // Create a pull consumer
        var consumer = await stream.CreateOrUpdateConsumerAsync(
            new ConsumerConfig("cons-fail") { AckPolicy = ConsumerConfigAckPolicy.Explicit },
            cts.Token);

        // Kill node 2
        await fixture.KillNode(2);
        await Task.Delay(2000, cts.Token);

        // Consumer should still work from node 0 (which is still alive)
        await using var client1 = fixture.CreateClient(1);
        await client1.ConnectAsync();
        var js1 = new NatsJSContext(client1);
        var stream1 = await js1.GetStreamAsync("JS_CONS_FAIL", cancellationToken: cts.Token);
        var cons1 = await stream1.GetConsumerAsync("cons-fail", cts.Token);
        cons1.ShouldNotBeNull();

        await fixture.RestartNode(2);
        await fixture.WaitForFullMeshAsync();
    }

    [Fact]
    public async Task R3Stream_Purge_ReplicatedAcrossNodes()
    {
        await using var client = fixture.CreateClient(0);
        await client.ConnectAsync();
        var js = new NatsJSContext(client);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(30));

        await js.CreateStreamAsync(
            new StreamConfig("JS_PURGE", ["js.purge.>"]) { NumReplicas = 3 },
            cts.Token);

        for (int i = 0; i < 5; i++)
            await js.PublishAsync("js.purge.data", $"msg-{i}", cancellationToken: cts.Token);

        await Task.Delay(1000, cts.Token);

        // Purge on node 0
        await js.PurgeStreamAsync("JS_PURGE", new StreamPurgeRequest(), cts.Token);

        await Task.Delay(1000, cts.Token);

        // All nodes should report 0 messages
        for (int node = 0; node < 3; node++)
        {
            await using var c = fixture.CreateClient(node);
            await c.ConnectAsync();
            var nodeJs = new NatsJSContext(c);
            var info = await nodeJs.GetStreamAsync("JS_PURGE", cancellationToken: cts.Token);
            info.Info.State.Messages.ShouldBe(0, $"Node {node} should have 0 messages after purge");
        }
    }
}
```

**Step 2: Run tests**

Run: `dotnet test tests/NATS.E2E.Cluster.Tests --filter "FullyQualifiedName~JetStreamCluster" -v normal`
Expected: All 4 PASS

**Step 3: Commit**

```bash
git add tests/NATS.E2E.Cluster.Tests/JetStreamClusterTests.cs
git commit -m "test: add JetStream cluster E2E tests (R3 replication, node failure, consumer failover, purge)"
```

---

### Task 9: Write GatewayFailoverTests (2 tests)

**Files:**
- Create: `tests/NATS.E2E.Cluster.Tests/GatewayFailoverTests.cs`

**Step 1: Write the tests**

```csharp
using NATS.Client.Core;
using NATS.E2E.Cluster.Tests.Infrastructure;

namespace NATS.E2E.Cluster.Tests;

public class GatewayFailoverTests(GatewayPairFixture fixture) : IClassFixture<GatewayPairFixture>
{
    [Fact]
    public async Task Gateway_Disconnect_Reconnects()
    {
        // Kill gateway B
        await fixture.KillNode(1);

        // Restart gateway B
        await fixture.RestartNode(1);
        await fixture.WaitForGatewayConnectionAsync();

        // Verify messaging resumes: pub on A, sub on B
        await using var clientA = fixture.CreateClientA();
        await using var clientB = fixture.CreateClientB();
        await clientA.ConnectAsync();
        await clientB.ConnectAsync();

        await using var sub = await clientB.SubscribeCoreAsync<string>("e2e.gw.reconnect");
        await clientB.PingAsync();

        // Allow interest propagation across gateway
        await Task.Delay(1000);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(10));
        await clientA.PublishAsync("e2e.gw.reconnect", "after-reconnect", cancellationToken: cts.Token);
        await clientA.PingAsync(cts.Token);

        var msg = await sub.Msgs.ReadAsync(cts.Token);
        msg.Data.ShouldBe("after-reconnect");
    }

    [Fact]
    public async Task Gateway_InterestUpdated_AfterReconnect()
    {
        // Subscribe on B, verify delivery from A
        await using var clientA1 = fixture.CreateClientA();
        await using var clientB1 = fixture.CreateClientB();
        await clientA1.ConnectAsync();
        await clientB1.ConnectAsync();

        await using var sub1 = await clientB1.SubscribeCoreAsync<string>("e2e.gw.interest");
        await clientB1.PingAsync();
        await Task.Delay(1000);

        using var cts1 = new CancellationTokenSource(TimeSpan.FromSeconds(10));
        await clientA1.PublishAsync("e2e.gw.interest", "before-kill", cancellationToken: cts1.Token);
        var msg1 = await sub1.Msgs.ReadAsync(cts1.Token);
        msg1.Data.ShouldBe("before-kill");

        // Kill B, restart, resubscribe
        await fixture.KillNode(1);
        await fixture.RestartNode(1);
        await fixture.WaitForGatewayConnectionAsync();

        await using var clientB2 = fixture.CreateClientB();
        await clientB2.ConnectAsync();
        await using var sub2 = await clientB2.SubscribeCoreAsync<string>("e2e.gw.interest");
        await clientB2.PingAsync();
        await Task.Delay(1000);

        using var cts2 = new CancellationTokenSource(TimeSpan.FromSeconds(10));
        await clientA1.PublishAsync("e2e.gw.interest", "after-restart", cancellationToken: cts2.Token);
        var msg2 = await sub2.Msgs.ReadAsync(cts2.Token);
        msg2.Data.ShouldBe("after-restart");
    }
}
```

**Step 2: Run tests**

Run: `dotnet test tests/NATS.E2E.Cluster.Tests --filter "FullyQualifiedName~GatewayFailover" -v normal`
Expected: All 2 PASS

**Step 3: Commit**

```bash
git add tests/NATS.E2E.Cluster.Tests/GatewayFailoverTests.cs
git commit -m "test: add gateway failover E2E tests (disconnect/reconnect, interest update)"
```

---

### Task 10: Write LeafNodeFailoverTests (2 tests)

**Files:**
- Create: `tests/NATS.E2E.Cluster.Tests/LeafNodeFailoverTests.cs`

**Step 1: Write the tests**

```csharp
using NATS.Client.Core;
using NATS.E2E.Cluster.Tests.Infrastructure;

namespace NATS.E2E.Cluster.Tests;

public class LeafNodeFailoverTests(HubLeafFixture fixture) : IClassFixture<HubLeafFixture>
{
    [Fact]
    public async Task Leaf_Disconnect_ReconnectsToHub()
    {
        // Kill leaf
        await fixture.KillNode(1);

        // Restart leaf
        await fixture.RestartNode(1);
        await fixture.WaitForLeafConnectionAsync();

        // Verify messaging: pub on hub, sub on leaf
        await using var hub = fixture.CreateHubClient();
        await using var leaf = fixture.CreateLeafClient();
        await hub.ConnectAsync();
        await leaf.ConnectAsync();

        await using var sub = await leaf.SubscribeCoreAsync<string>("e2e.leaf.reconnect");
        await leaf.PingAsync();
        await Task.Delay(1000);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(10));
        await hub.PublishAsync("e2e.leaf.reconnect", "leaf-back", cancellationToken: cts.Token);
        await hub.PingAsync(cts.Token);

        var msg = await sub.Msgs.ReadAsync(cts.Token);
        msg.Data.ShouldBe("leaf-back");
    }

    [Fact]
    public async Task Leaf_HubRestart_LeafReconnects()
    {
        // Kill hub
        await fixture.KillNode(0);

        // Restart hub
        await fixture.RestartNode(0);

        // Wait for leaf to reconnect (leaf has exponential backoff, may take a few seconds)
        await fixture.WaitForLeafConnectionAsync();

        // Verify messaging: pub on leaf, sub on hub
        await using var hub = fixture.CreateHubClient();
        await using var leaf = fixture.CreateLeafClient();
        await hub.ConnectAsync();
        await leaf.ConnectAsync();

        await using var sub = await hub.SubscribeCoreAsync<string>("e2e.leaf.hubrestart");
        await hub.PingAsync();
        await Task.Delay(1000);

        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(10));
        await leaf.PublishAsync("e2e.leaf.hubrestart", "hub-back", cancellationToken: cts.Token);
        await leaf.PingAsync(cts.Token);

        var msg = await sub.Msgs.ReadAsync(cts.Token);
        msg.Data.ShouldBe("hub-back");
    }
}
```

**Step 2: Run tests**

Run: `dotnet test tests/NATS.E2E.Cluster.Tests --filter "FullyQualifiedName~LeafNodeFailover" -v normal`
Expected: All 2 PASS

**Step 3: Commit**

```bash
git add tests/NATS.E2E.Cluster.Tests/LeafNodeFailoverTests.cs
git commit -m "test: add leaf node failover E2E tests (leaf restart, hub restart)"
```

---

### Task 11: Run full suite and final commit

**Step 1: Build the server host (E2E tests run against the compiled binary)**

Run: `dotnet build src/NATS.Server.Host`
Expected: Build succeeded

**Step 2: Run the full E2E cluster test suite**

Run: `dotnet test tests/NATS.E2E.Cluster.Tests -v normal`
Expected: 16 tests, all PASS

**Step 3: Run the original E2E suite to verify no regressions**

Run: `dotnet test tests/NATS.E2E.Tests -v normal`
Expected: 78 tests, all PASS

**Step 4: Final commit if any cleanup needed**

```bash
git add -A
git commit -m "test: finalize NATS.E2E.Cluster.Tests — 16 tests all passing"
```