Server becomes unresponsive to client access

Occasionally, a server becomes unresponsive to user access.  For example, clicking on a directory in Windows explorer may take several minutes to display the contents.

If ncpserver logging is set to debug (ncpcon log all), the following is frequently observed:
   ChangeSSGThread: Total number of threads active are XX and in queue are 1, spawning another thread!

Fixed in September 2015 OES 11 SP2 Scheduled Maintenance Update - 10853. See TID 7014420. Documented as:

A packet trace will help verify if you are encountering this issue.

• take a packet trace from the server or server's subnet (for at least 2 minutes)
• review the packet trace with the following filter (written in Wireshark notation):
  (ncp.type==0xbbbb && ncp.msg_flag == 0x24) || (ncp.func==87 && ncp.subfunc==34)
  
• in the filtered trace, you will see a number of "Server Broadcast Message" packets, but no reply.  For example, the following will be seen a number of times in a 2 minute trace:
  1696    server_IP    client_IP    NCP    C Server Broadcast Message
  1697    server_IP    client_ip    NCP    C Open CallBack Control (Op-Lock)
  1698    client_IP    server_ip    NCP    R Ok
  

If you are encountering this matter, a workaround is to:

1. Disable opportunistic locking (ncpcon set OPLOCK_SUPPORT_LEVEL = 0).
2. restart ndsd (or the entire host) to ensure all connections use the same opLock setting.

Note: users accessing data across a WAN link may see a slight decline in performance.