Specify fault domain

As there is a hard limit of 110 pods per node, I need to run multiple cluster nodes on the same vSphere host. Is there an option to define a fault domain that especially active datastores should not be located on the same bare metal?

Other option would be to separate master and worker nodes which is best practice in other distributions anyway.

The max pods is controlled by the kubelet. Why not just update the max pods per node?

Looking at microk8s documentation, the snap.microk8s.daemon-kubelet section in this doc explains where the arguments need to be.

It refers you to the kubelet documentation. There is a max-pods flag mentioned on the page.