How to troubleshoot excessive TCP Retransmission and TCP Dup Ack

[tachyon05]

I have 2 AV devices connected to the same access switch on the same VLAN.  The SVI for this VLAN is on the distribution switch, but I don't think either the SVI or the distribution switch is involved in this communication between the 2 devices.  These are AV devices and basically the issue is the [controller] 10.2.47.10 is not able to communicate with the [TV] 10.2.47.40 after some number of hours.  Power cycling or disconnect/reconnect the network cable on the TV resolves this issue for a number of hours until it happens again. This issue occurs in multiple rooms / floors / switches, but the setup is the same where the controller is on the same switch as the TV.  There is also evidence the controller is not the cause, because in all rooms that have this issue, controllers can control other TVs in the same room.

Packet capture on the access switch, at the time this issue is present, on both the TV port and Controller port show lots of TCP Dup Ack and Retransmission.  I think this suggest packet loss.  Any suggestions on what I can do on the network side to identify where the packet is lost?

[Pavel Tarakanov]

Are you sure that it's not the issue of the way this capture was collected?

If skip all these "retransmissions" and "dups" it seems to be normal TCP conversation between .10 and .40.

[tachyon05]

Would you please elaborate?  This is how the capture was configured on the access switch.  Other than the 2 ports, I think everything else is default.

Status Information for Capture av
Target Type:
Interface: GigabitEthernet1/0/42, Direction: BOTH
Interface: GigabitEthernet4/0/48, Direction: BOTH
Status : Inactive
Filter Details:
IPv4
Source IP: any
Destination IP: any
Protocol: any
Buffer Details:
Buffer Type: LINEAR (default)
Buffer Size (in MB): 10
File Details:
File not associated
Limit Details:
Number of Packets to capture: 0 (no limit)
Packet Capture duration: 0 (no limit)
Packet Size to capture: 0 (no limit)
Maximum number of packets to capture per second: 1000
Packet sampling rate: 0 (no sampling)

[MHM Cisco World]

Check output drop and show interface error count 

MHM

[Pavel Tarakanov]

If I understand correctly, 1/0/42 and 4/0/48 - ports where .10 and .40 connected. When packet arrived, for example on 1/0/42, it sent to capture. Then, when the same packet leave through 4/0/48, it again sent to the capture, because capture configured for both direction.

In you case it's more relevant to configure capture on one port in both direction or on both ports for one direction (ingress or egress).

Based on provided capture, I see that controller reach TV via telnet port. Do you know what kind of protocol used in this case? Is it telnet?

[tachyon05]

I see no input/output drops and interface errors do not increment.  Based on what I see, the amount of traffic is minimal, and mostly sending control signals, e.g., turn TV on/off, change volume, change input etc.