CPU/Memory limited leaf node using additional licenses

It seems when limiting a leaf node via numactl to a specific subset of memory & cpu, when calculating license usage the overall hardware available on the machine is used.

See the below commands used to bind instance to a specific numa node

-sh-4.2$ numactl -H
available: 4 nodes (0-3)
node 0 cpus: 0 1 2 3 4 5 6 7
node 0 size: 32767 MB
node 0 free: 31062 MB

sudo numactl --cpunodebind=0 --membind=0 memsqlctl start-node

-sh-4.2$ taskset -p -c 25600
pid 25600’s current affinity list: 0-7

*note 25600 is the memsql PID

If im not mistaken this should equate to 1 unit of usage however i’m seeing it counted as 4 in memsql studio as soon as the leaf node is started.

My overall goal is to run MemSQL within a small footprint on a larger box without using up too many license units so the instance can be properly clustered

Yes, licensing is based on the host, not the MemSQL process, so this is the expected behavior, as described in this FAQ:

https://docs.memsql.com/v7.0/introduction/faqs/memsql-faq/#what-are-license-units-and-how-do-they-apply-to-my-cluster

While constraining the resources based on numactl is unfortunately not supported by the licensing model, one way you can run MemSQL on a subset of the host’s resources is to deploy it within a VM. From the above link:

To artificially constrain your capacity, deploy MemSQL onto a virtual machine instead with fewer resources to lower your license unit count per host machine (which would be considered the virtual machine).