Incomplete discovery when trying to collect Nexus

[Alexandre Nascimento]

Hi,

We ran a collect in the CSPC 2.6 and the whole environment was rightly inventoried, but one Nexus is just being partially discovered. Apparently network, protocol and credentials are working, but we receive an error: "Incomplete discovery". What can we do to solve this issue?

Thanks.

[Lynden Price]

You can try to increase the timeout on it. Run a discovery on just the Nexus device, but on the last page before it starts, change the timeout from 3 seconds to 10 seconds etc. and see if it completes the discovery.

Thanks,

Lynden

[Alexandre Nascimento]

Thank you Lynden. In the past days a did some tests in timeout parameter, but the issue persists. We have 2x Nexus 7000 connected via VPC peerlink and the collector points to both management interface of each device. During these tests, I could see that sometimes a reach Nexus-1 or Nexus-2, but never reach both at the same time.

[Lynden Price]

Could you try running an SNMP walk from the command line of the CSPC and see if that times out as well?

The command is:

snmpwalk -v 2c -c <SNMPString> <IP Address>

That should give us a better idea of if it is a Collector problem or a connection issue.

Thanks,

Lynden

[Alexandre Nascimento]

Hi Lynden, we ran the command and received the attached error.

[Lynden Price]

Thanks Alexandre,

Is there a delay between the last successfull line on the snmpwalk output and the timeout? Does the walk hang for any amount of time or does it stop suddenly?

Thanks,

Lynden

[Alexandre Nascimento]

Lynden,

Answering your questions:

Is there a delay between the last successfull line on the snmpwalk output and the timeout?

Yes.

Does the walk hang for any amount of time or does it stop suddenly?

The walk hangs for any amount of time.

Additional information:

We discovered that sometimes SNMPWALK timeout in 40 seconds after the command execution, sometimes 100 seconds, 120 seconds. Well, the issue happens in different times.

We used the parameter -t 5000 (change timeout to 5 seconds) and SNMPWALK was concluded with success.

Trying to do the same thing in the collector, the issue persists. We set timeout up to 1200 seconds and the error happens before this time.

[Lynden Price]

The collector will not be able to inventory the device with that kind of delay. You'll need to determine if there is a routing issue causing the timeout or if there is simply too much load on the device to respond in the required amount of time.

[Alexandre Nascimento]

Hi Lynden,

We validated the route between the collector and the Nexus and the Nexus resources consumption and everything is right from these points of view. We could see that after the collector informs that the discovery was completed, the Nexus with issues in the collect keeps to send information about FEX, what does not happen with the Nexus that is rightly collected. I think that the problem is related with this, but I do not know how we can solve this. I am sending attached the dump file where we can see this behavior.

[Lynden Price]

This is unfortunately outside my area of expertise. Maybe you could cross post to the Nexus community? They should be able to interpret that output.

https://supportforums.cisco.com/community/4436/data-center