Potential memory leak

[DavidBoike]

Pretty simple repro, using NServiceBus 5.2.9 and NServiceBus.NHibernate 6.2.3 on .NET 4.6:

using System;
using System.Diagnostics;
using System.Threading;
using NServiceBus;
using NServiceBus.Persistence;
using NServiceBus.Config;
using NServiceBus.Config.ConfigurationSource;
using NServiceBus.Logging;

namespace TestForLeak3
{
    class Program
    {
        static void Main(string[] args)
        {
            LogManager.Use<DefaultFactory>().Level(LogLevel.Error);

            for (int run = 0; run < 5; run++)
            {
                Console.WriteLine($"Run {run+1}/5");
                for (int i = 0; i < 10; i++)
                {
                    using (var bus = GetBus())
                    {
                        Console.WriteLine($"  Bus {i + 1}/10");
                    }
                }
                Console.WriteLine("  Waiting 5s");
                Thread.Sleep(5000);
                Console.WriteLine("  Running GC");
                GC.Collect();
                Console.WriteLine("  GC done, waiting 5s, take snapshot now!");
                Thread.Sleep(5000);
            }
            Debugger.Break();
            Console.ReadLine();
        }

        private static IBus GetBus()
        {
            BusConfiguration cfg = new BusConfiguration();

            cfg.UseSerialization<XmlSerializer>();
            cfg.UsePersistence<NHibernatePersistence>()
                .ConnectionString(@"Data Source=.\SQLEXPRESS;Initial Catalog=FindMemLeak;Trusted_Connection=true;");
            cfg.EndpointName("MemLeakTest3");
            cfg.EnableInstallers();

            return Bus.Create(cfg).Start();
        }
    }

    class CustomConfig : IProvideConfiguration<MessageForwardingInCaseOfFaultConfig>,
            IProvideConfiguration<AuditConfig>
    {
        AuditConfig IProvideConfiguration<AuditConfig>.GetConfiguration()
        {
            return new AuditConfig
            {
                QueueName = "audit"
            };
        }

        public MessageForwardingInCaseOfFaultConfig GetConfiguration()
        {
            return new MessageForwardingInCaseOfFaultConfig
            {
                ErrorQueue = "error"
            };
        }
    }
}

Taking snapshots in VS2015 Diagnostic Tools yields a fairly predictable increase in heap objects with each run:

Although, this may be fairly low priority since it assumes bus instances are going to be created and destroyed constantly.

But, compare this to if you switch to InMemory persistence:

The sheer difference in scale may make this something to look at.

[SzymonPobiega]

Looking at the code, at least in V6 we don't ever dispose ISessionFactory. I suspect that it was the same in V5 and that this is the root cause of the leak. @Particular/nhibernate-persistence-maintainers should we prioritize this?

[DavidBoike]

I forgot I had opened this.

It only matters if you're continuously creating and destroying endpoints, which you really shouldn't do, right?

To me that gives it pretty low priority, the kind of change you'd roll into a minor you were going to release anyway if you had a free moment.