NDFC Service Restarting

[TerenceLockette]

Hello All,

I am completely new to the Nexus Dashboard product and am in the process of learning about it. I am currently using it in EVE-NG Pro where I have a lab setup using a single node ND running only the FC service. I'm using version 2.3.(2d) in my learning path but it appears that the NDFC service randomly restarts and I lose connectivity to the controller. Is this normal or unexpected behavior? I'm not sure if it's because ND requires a minimum of a 3-node cluster but it appears to be odd behavior. I'm considering using this for our organization to build out a VXLAN underlay but I'm hesitating because of what I'm seeing in my lab environment. Also, I have the proper min specs (16 vCPU, 64GB memory, and 500G storage with additional 550G of storage for the FC). For a bit more context, this is what I'm seeing consistently whether I'm actively working in ND or just logged in and sitting idle:

Any feedback or insight into this would be valuable.

Thanks!

[ADP89]

Hey Terence,

That behavior is not expected.

I have a few of comments on your setup though.

- ND[FC] is not technically supported on EVE-NG, meaning that we are not performing any sort of QA and/or testing on the solution.  I have seen that problem when running ND on non SSD disks.

- Support for a single vND node running FC was introduced on ND 3.0

- You are running a very old release of ND[FC]. 2 years old, many things changed and the application improved a lot in the meantime

In case you are looking for an existing lab where to test NDFC capabilities, may I suggest you to rely on dCloud labs? There is one named " Cisco Nexus Dashboard 3.2 - NDFC for VXLAN EVPN Multi-Site Deployments v1 " which offers a pre built topology where you can test many different options following a lab guide.

HTH,

ADP

[TerenceLockette]

Thank you for the feedback as it is greatly appreciated. I sort of figured that there would be some behavioral things I would encounter seeing it's in an emulated environment. However, I did add the newest 3.2.1i image to EVE and I noticed that once I began to import switches into the inventory is when things began to become unstable. I've been trying to see how the OOB ZTP function works and this is when I notice the issue. I haven't tested doing an inband ZTP or manually discovering/importing the switches and seeing if I get the same results. I may give dCloud a try but I find that some cloud-based labs are limited in some of the features or functionality that I want to test out. I was just curious as to whether this is something that is seen in production networks with a fully and properly deployed cluster.

[ADP89]

Understood. That is not something normally seen in production environments. The only times I saw something similar was when ND VMs where installed on top of normal HDDs and not SSDs.

If your host is using SSDs and hence the problem might be related to the nested virtualization why don't you deploy ND in a real VM(providing that you are in a vmware environment) and extend the network connectivity out of EVE-NG virtual topology? I do this every time in my lab (I use CML, not EVE but shouldn't change the final result), and works perfect

[TerenceLockette]

I will give this a try and see if the results are different for me. I am running vmware and have most of the physical resources allocated to my EVE-NG VM but I can allocate enough to a ND VM and extend it to EVE as suggested. Thanks for the advice. I'll see if I can get to it this week.