Container Service for Kubernetes:Enable CPU QoS for containers

Step 1: Apply the ConfigMap

1. Create a file named configmap.yaml with the following content:

   To apply CPU QoS to a workload such as a Deployment, add the koordinator.sh/qosClass label to the Pod template under template.metadata.

   apiVersion: v1
   kind: ConfigMap
   metadata:
     name: ack-slo-config
     namespace: kube-system
   data:
     # Enable the CPU QoS feature for containers.
     resource-qos-config: |
       {
         "clusterStrategy": {
           "lsClass": {
             "cpuQOS": {
               "enable": true,
               "groupIdentity": 2
             }
           },
           "beClass": {
             "cpuQOS": {
               "enable": true,
               "groupIdentity": -1
             }
           }
         }
       }

   The lsClass field configures LS Pods; beClass configures BE Pods. The following table describes the key parameters under cpuQOS.

   Parameter	Type	Valid values	Description
   enable	Boolean	true / false	true: enables CPU QoS cluster-wide. false: disables it.
   groupIdentity	Integer	-1 to 2	The kernel scheduling priority for the cgroup. Higher values mean higher priority. Default: 2 for LS Pods, -1 for BE Pods. Set to 0 to disable group identity for that class.

2. Check whether the ack-slo-config ConfigMap already exists in the kube-system namespace:

   • If it exists, run the following patch command to update it without overwriting other configuration fields:

     kubectl patch cm -n kube-system ack-slo-config --patch "$(cat configmap.yaml)"

   • If it does not exist, create it:

     kubectl apply -f configmap.yaml

Step 2: Deploy an LS Pod and verify

1. Create a file named ls-pod-demo.yaml with the following content:

   apiVersion: v1
   kind: Pod
   metadata:
     name: ls-pod-demo
     labels:
       koordinator.sh/qosClass: 'LS' # Specify the QoS class of the Pod as LS.
   spec:
     containers:
     - command:
       - httpd
       - -D
       - FOREGROUND
       image: registry.cn-zhangjiakou.aliyuncs.com/acs/apache-2-4-51-for-slo-test:v0.1
       imagePullPolicy: Always
       name: apache
       resources:
         limits:
           cpu: "4"
           memory: 10Gi
         requests:
           cpu: "4"
           memory: 10Gi
     restartPolicy: Never
     schedulerName: default-scheduler

2. Deploy the Pod:

   kubectl apply -f ls-pod-demo.yaml

3. On the node, check the group identity assigned to the LS Pod's cgroup:

   cat /sys/fs/cgroup/cpu/kubepods.slice/kubepods-pod1c20f2ad****.slice/cpu.bvt_warp_ns

   Expected output:

   # The group identity of the LS Pod is 2, which indicates high priority.
   2

Step 3: Deploy a BE Pod and verify

1. Create a file named be-pod-demo.yaml with the following content:

   apiVersion: v1
   kind: Pod
   metadata:
     name: be-pod-demo
     labels:
       koordinator.sh/qosClass: 'BE' # Specify the QoS class of the Pod as BE.
   spec:
     containers:
       - args:
           - '-c'
           - '1'
           - '--vm'
           - '1'
         command:
           - stress
         image: registry-cn-beijing.ack.aliyuncs.com/acs/stress:v1.0.4
         imagePullPolicy: Always
         name: stress
     restartPolicy: Always
     schedulerName: default-scheduler

2. Deploy the Pod:

   kubectl apply -f be-pod-demo.yaml

3. On the node, check the group identity assigned to the BE Pod's cgroup:

   cat /sys/fs/cgroup/cpu/kubepods.slice/kubepods-besteffort.slice/kubepods-besteffort-pod4b6e96c8****.slice/cpu.bvt_warp_ns

   Expected output:

   # The group identity of the BE Pod is -1, which indicates low priority.
   -1

The LS Pod has a group identity of 2 and the BE Pod has a group identity of -1. This difference confirms that CPU QoS is active: the kernel prioritizes CPU resources for LS workloads over BE workloads.

Protocol compatibility when upgrading to ack-koordinator

Earlier versions of ack-slo-manager (v0.8.0 and earlier) used the alibabacloud.com/qosClass annotation to configure CPU QoS. ack-koordinator supports both protocols, so you can upgrade the component and migrate your Pods to the koordinator.sh protocol incrementally. Note that support for the older alibabacloud.com protocol ended on July 30, 2023—update your resources to the new protocol as soon as possible.