we use RKE for setting up the cluster with 4 master node (VM, 8 core CPUcurrentlysourcesstagingahorriblea workers and 16G ram) and 5 worker node(bare-metal, HPE G10, 88 core CPU, 256G ram)
Kubernetes version:1.15
Host OS: coreos
and increase max pod per node to 250 and currently, we have about 500-600 total pods.
most of our stack is php-fpm.
yesterday we run a load test with autocannon from multiple sources on our production (staging namespace 
) and yes I know its horrible but we think cluster is ok and tolerate it due to our monitoring background history.
I want to know do you recommend that we use many workers (VM) instead of few workers (bare-metal)?