How to properly choose cpu and memory for different k8s pods?

[zhangyuqiu]

I notice that when creating a graph scope session, there will be configs that like these:

"k8s_coordinator_cpu",
"k8s_coordinator_mem",
"k8s_etcd_num_pods",
"k8s_etcd_cpu",
"k8s_etcd_mem",
"k8s_vineyard_daemonset",
"k8s_vineyard_cpu",
"k8s_vineyard_mem",
"vineyard_shared_mem",
"k8s_engine_cpu",
"k8s_engine_mem",
"k8s_mars_worker_cpu",
"k8s_mars_worker_mem",
"k8s_mars_scheduler_cpu",
"k8s_mars_scheduler_mem",

Is there a guide for how to properly assign resources for these pods? Maybe not precise, but can give a rough picture of how much or what percentage of system resource should a pod takes. Thanks!