psipro_1989
New Around Here
Hello
I hope someone here will have suggestions for me. I have several small network attached data servers providing periodic UDP packets at constant rates (71 bytes at 10 Hz). The servers are connected 10 T and reside in a custom switch rack with a Gb uplink to a Cisco 2960G and from there to a Windows client. An active TCP socket must be maintained or the between the client and the server of the server will reset its internal application, stopping the data stream.
This configuration has run for years with a WinXP client but now with the requirement to change to Win7 I'm having problems.
With the client application running on Win7, one of the servers will intermittently stop responding to TCP ACKs and the client will (after the appropriate number of re-transmissions) send a fin packet. The server obligingly resets itself and announces its availability for connection.
The funny thing is that the application in the server continues to send UDP data even though the TCP keepalive ACKS are un acknowledged. It continues to believe that the connection is valid.
If the client application is run on a WinXP machine then the "disconnect" (due to missing keepalives) doesn't occur.
IP6 is disabled on all network hosts.
Question:
Is there something in Win7 (vs WinXP) that would cause a switch to stop transmitting TCP keepalive acks?
Has anyone experienced something similar, where a switch apparently stops transmitting TCP ACKs for a socket where keepalive ACK packets are the only traffic?
I also have one reported incidence of an HP Procurve involved in the same symptom at a different site. A reset (power off) of that switch resolved the issue and it has not reoccurred.
I feel as if this might be an ARP table issue in the switch but I can't think of how to test this theory. I don't have access to the Cisco or Procurve CLI and insufficient knowledge of the switch internals to be able to formulate a test case.
C.
I hope someone here will have suggestions for me. I have several small network attached data servers providing periodic UDP packets at constant rates (71 bytes at 10 Hz). The servers are connected 10 T and reside in a custom switch rack with a Gb uplink to a Cisco 2960G and from there to a Windows client. An active TCP socket must be maintained or the between the client and the server of the server will reset its internal application, stopping the data stream.
This configuration has run for years with a WinXP client but now with the requirement to change to Win7 I'm having problems.
With the client application running on Win7, one of the servers will intermittently stop responding to TCP ACKs and the client will (after the appropriate number of re-transmissions) send a fin packet. The server obligingly resets itself and announces its availability for connection.
The funny thing is that the application in the server continues to send UDP data even though the TCP keepalive ACKS are un acknowledged. It continues to believe that the connection is valid.
If the client application is run on a WinXP machine then the "disconnect" (due to missing keepalives) doesn't occur.
IP6 is disabled on all network hosts.
Question:
Is there something in Win7 (vs WinXP) that would cause a switch to stop transmitting TCP keepalive acks?
Has anyone experienced something similar, where a switch apparently stops transmitting TCP ACKs for a socket where keepalive ACK packets are the only traffic?
I also have one reported incidence of an HP Procurve involved in the same symptom at a different site. A reset (power off) of that switch resolved the issue and it has not reoccurred.
I feel as if this might be an ARP table issue in the switch but I can't think of how to test this theory. I don't have access to the Cisco or Procurve CLI and insufficient knowledge of the switch internals to be able to formulate a test case.
C.