Aaron Friesen
2016-11-08 17:07:39 UTC
Thanks in advance for assistance.
Using ZeroMQ C# bindings (4.1.0.21) with libzmq (4.1.5.0) on Windows (7, Server 2008 R2, and Server 2012).
For request/response functionality I am using a single Router socket (tcp) in each of several interconnected processes.
Each process has an identifier (GUID) that is known by all other processes, which is used as the Identity on the socket.
Each process performs a Bind on a specific port (also known by all other processes), and also performs a Connect to each of the other Router sockets in the other processes.
For example, if 5 processes are used, visualize a completely interconnected star.
When all 5 processes are initially started, communication works great. I can send and receive messages between any of the 5.
The problem comes if for some reason I must restart one of the 5. Once that process is restarted, communication with the other 4 seems to go nowhere, for both incoming and outgoing traffic.
I have monitors on all of the sockets which seems to show that everything is reconnecting.
I also have RouterMandatory set to Report. Sends from the newly restarted process do not Report when attempting to Send to the other processes, and the other processes do not Report when attempting to Send to the restarted process.
If I stop one of the other remaining processes, then attempts to Send to that process Report as expected.
I also have pub/sub socket connections, with each process publishing a heartbeat, and those sockets all successfully reconnect.
I am testing all of this from (in this example) a 6th process that simply connects to the other 5 processes (also using a Router socket, no Bind involved), and it successfully reconnects to the restarted process without any problems and can send messages (and receive subsequent response) with any of the other 5, including any processes that are restarted. This 6th process does not publish to the other 5, but does subscribe to the heartbeats. I am convinced that the fact that 6th process successfully reconnects is a clue to what I am doing wrong, but it has been insufficient for me to find out what the root problem really is.
The only other thing to note is that Linger is set to zero prior to Bind/Connect.
The code for all processes for setting up the Router socket is:
using (_requestSocket = new ZSocket(_context, ZSocketType.ROUTER))
{
try
{
_requestSocket.Monitor(Endpoint_RequestMonitor, out error);
Debug.Assert(error == ZError.None);
_requestSocket.Identity = Encoding.UTF8.GetBytes(_selfID);
_requestSocket.Linger = TimeSpan.Zero;
_requestSocket.RouterMandatory = RouterMandatory.Report;
_requestSocket.Bind(Endpoint_RequestInproc, out error);
Debug.Assert(error == ZError.None);
endpoint = string.Format("tcp://*:{0}", _requestPort);
_requestSocket.Bind(endpoint, out error);
Debug.Assert(error == ZError.None);
}
catch (Exception ex)
{
Environment.Exit(-1);
}
Thanks for any assistance or advice you can provide,
Aaron
Using ZeroMQ C# bindings (4.1.0.21) with libzmq (4.1.5.0) on Windows (7, Server 2008 R2, and Server 2012).
For request/response functionality I am using a single Router socket (tcp) in each of several interconnected processes.
Each process has an identifier (GUID) that is known by all other processes, which is used as the Identity on the socket.
Each process performs a Bind on a specific port (also known by all other processes), and also performs a Connect to each of the other Router sockets in the other processes.
For example, if 5 processes are used, visualize a completely interconnected star.
When all 5 processes are initially started, communication works great. I can send and receive messages between any of the 5.
The problem comes if for some reason I must restart one of the 5. Once that process is restarted, communication with the other 4 seems to go nowhere, for both incoming and outgoing traffic.
I have monitors on all of the sockets which seems to show that everything is reconnecting.
I also have RouterMandatory set to Report. Sends from the newly restarted process do not Report when attempting to Send to the other processes, and the other processes do not Report when attempting to Send to the restarted process.
If I stop one of the other remaining processes, then attempts to Send to that process Report as expected.
I also have pub/sub socket connections, with each process publishing a heartbeat, and those sockets all successfully reconnect.
I am testing all of this from (in this example) a 6th process that simply connects to the other 5 processes (also using a Router socket, no Bind involved), and it successfully reconnects to the restarted process without any problems and can send messages (and receive subsequent response) with any of the other 5, including any processes that are restarted. This 6th process does not publish to the other 5, but does subscribe to the heartbeats. I am convinced that the fact that 6th process successfully reconnects is a clue to what I am doing wrong, but it has been insufficient for me to find out what the root problem really is.
The only other thing to note is that Linger is set to zero prior to Bind/Connect.
The code for all processes for setting up the Router socket is:
using (_requestSocket = new ZSocket(_context, ZSocketType.ROUTER))
{
try
{
_requestSocket.Monitor(Endpoint_RequestMonitor, out error);
Debug.Assert(error == ZError.None);
_requestSocket.Identity = Encoding.UTF8.GetBytes(_selfID);
_requestSocket.Linger = TimeSpan.Zero;
_requestSocket.RouterMandatory = RouterMandatory.Report;
_requestSocket.Bind(Endpoint_RequestInproc, out error);
Debug.Assert(error == ZError.None);
endpoint = string.Format("tcp://*:{0}", _requestPort);
_requestSocket.Bind(endpoint, out error);
Debug.Assert(error == ZError.None);
}
catch (Exception ex)
{
Environment.Exit(-1);
}
Thanks for any assistance or advice you can provide,
Aaron