resource issue when using sdwan-lab-deployment-tool in CML 2.8

[thibaultm]

 

Hello,

I have a powerful laptop with CML 2.8 installed.
When I stumbled upon the video : Build your own Cisco Catalyst SD-WAN Virtual Lab with free Lab Deployment Tool in CML - YouTube and another one : Easy SD-WAN in CML with catalyst-sdwan-lab -
with the same Python script :
sdwan-lab-deployment-tool/README.md at main · cisco-open/sdwan-lab-deployment-tool · GitHub
So I tested it.

I have a VM for CML 2.8 and a VM for Network automation with ubuntu 22.04.
I installed the script on my ubuntu 22.04 VM and followed the instructions of both youtube videos.
Everything went OK until I started the deploy task :

(sdwan_lab1) thibault@NetAutoU2204:~/csdwan$ vi sdwan_lab1.sh
(sdwan_lab1) thibault@NetAutoU2204:~/csdwan$ source sdwan_lab1.sh
(sdwan_lab1) thibault@NetAutoU2204:~/csdwan$ csdwan setup
Setup task done
(sdwan_lab1) thibault@NetAutoU2204:~/csdwan$ csdwan deploy 20.15.1
But it can to connect to the vManage controller.
Connected to the lab everything is started except for the server Manager01.
and when I try to start it manualy it fails with the message :
"Failed to choose a suitable compute host (cml-controller has insufficient total HW resources.)"
which is surprising because my CML VM has 32GB of RAM and 8 cores of CPU.

After having googled a bit I stumbled on this video by Joe clark explaining a similar issue :
CML: Increase the Resources of a Running Node
But in my case Manager01 never started (see screenshot above) I think because I cannot edit the document like he does in the video.
I doubled the amount of RAM of my VM to 64 GB : same result, I don't know what to do I am stuck.
That could be a generic issue with default sizing of Manager01 in this lab but of course I can't be sure of that.

[Ramblin Tech · Accepted solution]

From the FAQ in the CML docs:

8. What resources are required by each node type?

The following table lists the nominal vCPU, memory, and disk size for each node type. The resource usage of a particular node in a lab simulation depends on the node type, selected VM image, and the associated configuration. Note that CML over-provisions your system's actual hardware for virtualization, and CML uses Linux kernel same page merging to reduce the memory required to run a lab, especially when the lab has multiple nodes of the same type. Therefore, the resources required to run a simulation will generally be less than the nominal resource requirements of all of its nodes.

Cisco Reference Platforms:

Node Type	vCPUs	Memory	Minimum Disk Consumption	Maximum Disk Consumption
SD-WAN Edge	2	5 GB	500 MB	8 GB
SD-WAN Controller	2	4 GB	500 MB	10.5 GB
SD-WAN Manager	8	16 GB	500 MB	277 GB
SD-WAN Validator	2	4 GB	500 MB	10.5 GB
SD-WAN vEdge	2	2 GB	500 MB	10.5 GB

Your other nodes (non-SDWAN) will consume CPU and RAM resources as well (I only pasted the SD-WAN components above). At the botton of the Workbench view in CML, you will see some resource consumption indicators. What resources are nearing depletion? Note that CML will not allow you to start a new node when RAM consumption is greater than 80%.

[thibaultm]

Thanks.
OK yesterday evening I had to stop the CML VM and of course the lab.
So this morning with the lab all shut, I have the following figures at the bottom of the workbench :
CPU from 0.60% to 1.4 %
Memory 1.12 %
Disk 58.70 %

With everything else stopped I tried to launch Manger01 alone, and it does not even start with all my VM resource just for it :

 

[thibaultm]

Now I must admit my laptop is a "beast" my VM was set up with 8 processors 1 cores, I changed it for 8 processors 2 cores and it starts GREAT !
Thanks for the reply (I know I should have RTFM...)
Thanks for those who made CML and the script to start such interesting labs !
I will have to try to creates edges now of course...