KVM Connections through KVM Manager interface using wrong IP

[ashley.smith1]

Hello everyone.  I have a very odd problem happening.  I've configured my CIMC interfaces with an inbound IP address.  When I try to connect to the KVM console from the UCS Manager, everything works correctly.  However, as soon as I try to use the KVM Manager interface to connect to the KVM console, the IP used is not the one which is assigned to the CIMC, but the subnet ID instead (in my case 10.3.3.0:2068).  Has anyone else noticed this behaviour?  It's the strangest thing.  As a side note, I've also tried using an Outband IP as well, and I get the same results.

Thanks for any help!

Ashley

[Clifford Aldan]

Ashley, 

Can you provide some screenshots of what you are observing? 

What's the UCS firmware? 

Does the KVM console actually launch? 

[ashley.smith1]

Hi Clifford,

Sorry for my delaying in replying, and thank you for responding to my question.  We are currently running the firmware version 2.2.5(b).  The console never successfully launches, despite the Java application launching and trying to connect.  Below are the screenshots.

1. Selection of the IP to use for the KVM
   
2. Netstat of all connections to that subnet, showing that the destination actually used for the KVM connection is the 10.3.3.0 on the port 2068
   
3. Process Java that corresponds to the PID 18232
   
4. Connection failure message
   

Thank you again.  Please let me know if there is any other information I may be able to provide you.

[Wes Austin]

Hey Ashley,

Can you confirm if the IP address you are assigning is in the same subnet as the mgmt interfaces on the FI? This is required for connectivity.

http://www.cisco.com/c/en/us/support/docs/servers-unified-computing/ucs-manager/115958-ucsm-kvm-troubleshooting-00.html

1. If you are unable to ping the KVM IP address, which was assigned from the management (MGMT) IP pool, you must verify that the IP pool addresses are within the same subnet as the management IP address assigned to the fabric interconnect. If the pool does not match, all KVMs that receive IP addresses from this pool are impacted.

2. Verify TCP port 2068 is not blocked by the access control list (ACL) or firewall between client and blade CIMC. With this port closed, it results in failure to connect to any UCS KVM.

3. Verify browser configuration to ensure HTTP proxy does not break communication.

4. If you are unable to ping the KVM IP address, check which fabric interconnect manages the instance for the blade and verify if it responds. If a fabric interconnect suffers from a failed/down mgmt0 interface, all blades that are managed by it experience KVM accessibility issues.

HTH,

Wes

[ashley.smith1]

Hi Wes,

This is an in-band CIMC address, which is connected to a VLAN which is trunked to the correct uplink interfaces.  There are no network connectivity problems, I can ping the interfaces without a problem, and when the KVM is launched from the UCS System Manager console the KVM connects without any problems.  Note that the problem also occurs with any out-of-band interfaces as well.

It's the very illogical nature of the IP address that the KVM Manager tries to connect to which made me ask this question.  It essential renders the KVM Manager useless for us.

Here the pings pass without any trouble:

Here is a telnet connection to port 2068 which succeeds:

Here is the telnet PID:

Thank you for you continued help.