Hey just a few clarifications.
For all discussion below I’m using n1-highcpu-64 which is defined as:
- “High-CPU machine type with 64 vCPUs and 57.6 GB of memory.”
- “For the n1 series of machine types, a vCPU is implemented as a single hardware hyper-thread”
- I notice you did
+S 128with6400mbut above you used+S 24with12000m. Is it supposed to be1000 * # of cores / 2or100 * # of cores / 2 - I’m a little confused on the teriminology of “cores”. My server has 64 vCPU from hyperthreading. I assumed that means I have 64 cores, 64 vCPUs, and 32 “actual” hardware CPUs. Is that correct?
- You mention leaving a little extra space for KubeDNS. Do I set those limits for KubeDNS by limiting the resources for my app? Or is there a way to set it ‘directly’? Can I have you confirm these numbers for sanity on a single node specified above?
number of nodes
1 n1-highcpu-64
number of pods
1
imaginary number of CPUs allocated to KubeDNS
2
vm.args
+S 124
pod spec (not sure on multiplier)
limits:
cpu: 62000m
requests:
cpu: 62000m
Edit cpu: 62000m just defaults to cpu: "62" in GKE so I’m sure that’s right. Now my only question… do I allow more CPU’s for KubeDNS by doing it this way or do I specify explicitly for KubeDNS elsewhere and leave this at "64" instead of “62”`
