After your SCADA node runs for a little while, iFIX clients start to loose connections. Eventually TCPTASK crashes. When you examine the iFIX daily event log you see TCPTASK entries of the form "[TCPTASK.EXE] TCP/IP: Error from accept(). SOCKERR = 10055, fixerr = 8515".

Resolution

The fix error 8515 maps to "TCPTASK: Windows Socket Library resource problems". The error comes from our TCPTASK code that listens for new connections. In this code there is a While loop that continuously checks for new incoming connections over the listen socket (2010). When a client attempts a connection, we attempt to accept the connection with accept(). After the accept() call we note any errors. If there is one, we call WSAGetLastError(). This is what returns WINSOCK error 10055. This WINSOCK error indicates a shortage of resources on your system. In this situation, TCPTASK cannot create a new socket for the new connection.

Using Windows Performance Monitor and Microsoft Poolmon, check to see where you are leaking pool memory. More than likely you will find a low level driver using the non-paged pool that is leaking kernel objects (a socket is a type of kernel object).

References

http://support.microsoft.com/default.aspx?scid=kb;EN-US;q177415    "How to use Memory Pool Monitor (Poolmon.exe) to troubleshoot kernel mode memory leaks"

http://blogs.msdn.com/ntdebugging/archive/2006/12/18/Understanding-Pool-Consumption-and-Event-ID_3A00_--2020-or-2019.aspx  "Understanding Pool Consumption and Event ID: 2020 or 2019"