System.InvalidOperationException: The SessionId was not set on a message on wolverine-ping when sending to an Azure FIFO queue

[s11ha]

Describe the bug
While wolverine application is running when queue is latched then in order to resume wolverine sends a "wolverine-ping" envelope.
But if queue is Azure FIFO queue it fails with an error

System.InvalidOperationException: The SessionId was not set on a message, and it cannot be sent to the entity. 
Entities that have session support enabled can only receive messages that have the SessionId set to a valid value.

then it enters into a cycle of being latched, trying to resume, getting this error until app is stopped.

The same error happens when FailureAcknowledgement is being sent. But for FailureAcknowledgement i guess it could be eliminated by using AddOutgoingRule for a FIFO queue wolverine configuration.
I guess the same could happen and with Acknowledgement but I never encountered that.

To Reproduce
I prepared a demo app where a wolverine-ping is simulated because I do not know how to cause it other way.

public async Task<Response> Handle(
    Request request,
    IConfiguration configuration,
    AzureServiceBusTransport azureServiceBusTransport,
    IWolverineRuntime runtime,
    IMessageBus messageBus,
    CancellationToken token
)
{
    var queue = azureServiceBusTransport.Queues[configuration["AsbFifoQueue"]];
    var methodInfo = queue
        .GetType()
        .GetMethod("CreateSender", BindingFlags.Instance | BindingFlags.NonPublic);
    var sender = (ISender)methodInfo.Invoke(queue, [runtime]);
    var sendingAgent = runtime.Endpoints.CreateSendingAgent(
        new Uri("local://replies/"),
        sender,
        queue
    );
    ((ISenderRequiresCallback)sender).RegisterCallback(sendingAgent as ISenderCallback);

    var pingResult = await sender.PingAsync();

    return new Response(pingResult);
}

The full code is here - PingCheck.zip

Steps to reproduce the behavior:

1. open appsettings.json and specify AsbConnectionString and AsbFifoQueue (name of fifo queue)
2. dotnet run
3. navigate to http://localhost:5000/ping
4. See error in the loop
5. you also can notice the PingAsync always returns true

Expected behavior
There should not be error

Desktop (please complete the following information):

• Version [3.6.5]

Additional context

The hotfix is to use Harmony patching library.

[HarmonyPatch(typeof(Envelope), nameof(Envelope.ForPing))]
public class EnvelopeForPingFix
{
    public static void Postfix(ref Envelope __result, Uri destination)
    {
        __result.GroupId = Envelope.PingMessageType;
    }
}

public static class WolverinePingFix
{
    [ModuleInitializer]
    public static void Patch()
    {
        var harmony = new Harmony(nameof(WolverinePingFix));
        var assembly = Assembly.GetExecutingAssembly();
        harmony.PatchAll(assembly);
    }
}

Maybe it can be fixed the same way in wolverine.

public static Envelope ForPing(Uri destination)
{
        return new Envelope
        {
            MessageType = PingMessageType,
            // fix
            GroupId = Envelope.PingMessageType,
            // end fix
            Data = [1, 2, 3, 4],
            ContentType = "wolverine/ping",
            Destination = destination
        };
}

[jeremydmiller]

Thanks for the write up here. It's probably going to take less time to do the fix than it did your issue, but only because you did write this up.