numa

https://docs.openshift.com/container-platform/4.3/scalability_and_performance/using-topology-manager.html#topology_manager_policies_using-topology-manager

https://www.sharcnet.ca/help/index.php/Using_numactl

video

• https://youtu.be/J2VQQZxk3eY
• https://www.bilibili.com/video/BV1HK4y1r7Di/

oc get featuregate/cluster -o yaml

oc patch featuregate/cluster -p '{"spec": { "featureSet": "LatencySensitive" } }' --type=merge

oc get KubeletConfig -o yaml

cat << EOF > cpumanager-kubeletconfig.yaml
apiVersion: machineconfiguration.openshift.io/v1
kind: KubeletConfig
metadata:
  name: cpumanager-enabled
spec:
  machineConfigPoolSelector:
    matchLabels:
      custom-kubelet: cpumanager-enabled
  kubeletConfig:
     cpuManagerPolicy: static
     cpuManagerReconcilePeriod: 5s
     topologyManagerPolicy: single-numa-node 
EOF
oc apply -f cpumanager-kubeletconfig.yaml

oc project demo 

cat << EOF > cpumanager-pod.yaml
apiVersion: v1
kind: Pod
metadata:
  generateName: cpumanager-
spec:
  containers:
  - name: cpumanager
    image: gcr.io/google_containers/pause-amd64:3.0
    resources:
      requests:
        cpu: 1
        memory: "1G"
      limits:
        cpu: 1
        memory: "1G"
  nodeSelector:
    cpumanager: "true"
EOF
oc apply -f cpumanager-pod.yaml

# on the worker node
yum install numactl
# 指定命令运行在NUMA NODE0上（CPU，内存都来自NUMA NODE0）
numactl --cpunodebind=0 --membind=0 COMMAND
# 指定命令CPU来自NUMA NODE1，内存尽可能来自NUMA NODE1，如果NUMA NODE1没有足够的内存了，则使用NUMA NODE0上的内存
numactl --cpunodebind=1 --preferred=1 COMMAND
# 获取进程cpu的mask
taskset -p <pid>
# pid 26624's current affinity mask: ff  这个是没设置掩码

# 进程的memory信息可以通过命令获取
numastat <pid>
# Per-node process memory usage (in MBs) for PID 26624 (firefox)
#                            Node 0           Total
#                   --------------- ---------------
# Huge                         0.00            0.00
# Heap                         0.00            0.00
# Stack                        0.08            0.08
# Private                    208.50          208.50
# ----------------  --------------- ---------------
# Total                      208.58          208.58
# 类似于进程，在某个NUMA Node上占用多少内存

# 查询PCI网卡设备所在numa node
cat /sys/class/net/<devicename>/device/numa_node


# back to normal
cat << EOF > cpumanager-kubeletconfig.yaml
apiVersion: machineconfiguration.openshift.io/v1
kind: KubeletConfig
metadata:
  name: cpumanager-enabled
spec:
  machineConfigPoolSelector:
    matchLabels:
      custom-kubelet: cpumanager-enabled
  kubeletConfig:
     cpuManagerPolicy: static
     cpuManagerReconcilePeriod: 5s
     topologyManagerPolicy: none 
EOF
oc apply -f cpumanager-kubeletconfig.yaml

# delete them all
oc delete -f cpumanager-kubeletconfig.yaml