5

I'm using NetMQ for client/server comms for a pricing system using Pub/Sub for price data and Req/Res for C&C. It's working very well except that I'm seeing random and unreproducible exceptions occurring in the bowels of the NetMQ library.

I'm currently using NetMQ v4.0.1.6 on a .Net v4.7.2 stack on Windows Server 2008R2/20012R2.

The exceptions include:

  • NetMQ.FaultException @ NetMQ.Msg.Close()
  • System.ArgumentOutOfRangeException @ NetMQ.Core.Patterns.Utils.ArrayExtensions.Swap[]
  • System.IndexOutOfRangeException @ NetMQ.Core.Utils.YQueue`1[]/NetMQ.Core.Patterns.Utils.Trie.Check()
  • System.Net.Sockets.SocketException @ NetMQ.Core.Mailbox.TryRecv()/NetMQ.Core.Mailbox.Send()
  • System.NullReferenceException @ NetMQ.Core.Patterns.Utils.Trie.ApplyHelper()/NetMQ.Core.Patterns.Utils.Trie.Check()
  • System.ObjectDisposedException @ NetMQ.Core.SocketBase.CheckDisposed()

and are all happening within either the NetMQ.NetMQPoller.RunPoller() or NetMQ.Core.Utils.Proactor.Loop() and I'm at a loss as to how to catch and mitigate them.

What is the correct approach to harden a service/app using NetMQ to prevent such exceptions bringing down the service?

SteveH
  • 343
  • 2
  • 10
  • I can add one more exception to the list: Unhandled exception: System.ArgumentNullException: Value cannot be null. Parameter name: remoteEP at System.Net.Sockets.Socket.SendTo(ByteD buffer, Int32 offset, 1nt32 size, SocketFlags socketFlags, EndPoint remoteEP) at NetN1Q.NetNIQBeacon.Shim.SendUdpFrame(NetMO.Frame frame) at NettVIQ.NetMQSocket.InvokeEvents(Object sender, PollEvents events) at NetMQ.NetMQPoller.RunPoller() at NetMQ.NetMQPoller.Run(SynchronizationContext syncContext) – Dimka Mar 20 '21 at 14:25
  • Is your MQ queues only accessible from an internal network ? is there any ACL on the queues ? – Zilog80 Mar 26 '21 at 16:12

1 Answers1

-1

I did this professionally, and with great success, at one large organisation.

The best practice is to include within your loop a try catch handler, and then do something like follows

(Example error types plucked from your list at random)

/*loop*/(/*loop condition*/) {
  try {
    // processing
  } 
  catch (NetMQ.FaultException $exception) {
    // specific error handling
  } 
  catch (System.ObjectDisposedException $exception) {
    // specific error handling
  }
  catch (Exception $exception) {
    // generic error handling
  }
}

Also, feel free to pass it a specific error handling class/service to offload your error handling functionality into a separate location (greater testability).

JohnFF
  • 723
  • 5
  • 20
  • 1
    I believe it won't help much because those seems to be uncatched exceptions thrown by the library code from another thread. And even catching those won't prevent program from terminating. – Dimka Mar 27 '21 at 14:34