没有根据节点实际负载调度Pod

crane版本：helm scheduler-0.2.2 k8s版本：1.24

使用一台k8s节点，规格16核32GB 节点Annotations负载

Annotations:        alpha.kubernetes.io/provided-node-ip: 172.30.64.34
                    cpu_usage_avg_5m: 0.63012,2023-10-17T15:04:32Z
                    cpu_usage_max_avg_1d: 0.63666,2023-10-17T14:03:36Z
                    cpu_usage_max_avg_1h: 0.63654,2023-10-17T15:01:29Z
                    mem_usage_avg_5m: 0.21519,2023-10-17T15:04:34Z
                    mem_usage_max_avg_1d: 0.21614,2023-10-17T14:02:41Z
                    mem_usage_max_avg_1h: 0.21700,2023-10-17T15:01:53Z
                    node.alpha.kubernetes.io/ttl: 0
                    node_hot_value: 0,2023-10-17T15:04:34Z

节点Requests

Allocated resources:
  (Total limits may be over 100 percent, i.e., overcommitted.)
  Resource                    Requests       Limits
  --------                    --------       ------
  cpu                         15367m (96%)   1770m (11%)
  memory                      25943Mi (91%)  1500Mi (5%)
  ephemeral-storage           0 (0%)         0 (0%)
  hugepages-1Gi               0 (0%)         0 (0%)
  hugepages-2Mi               0 (0%)         0 (0%)
  attachable-volumes-aws-ebs  0              0

创建一个测试服务，共有8个pod副本，每个pod压测产生2核1GB负载，requests为3核5GB

apiVersion: apps/v1
kind: Deployment
metadata:
  name: demo-nginx
  namespace: demo
  labels:
    app: demo-nginx
spec:
  replicas: 8
  selector:
    matchLabels:
      app: demo-nginx
  template:
    metadata:
      labels:
        app: demo-nginx
    spec:
      schedulerName: crane-scheduler
      containers:
      - name: demo-nginx
        image: xxxxxx/stress:latest
        command: ["stress", "-c", "1","--vm", "1", "--vm-bytes", "1G", "--vm-keep"]
        ports:
        - containerPort: 80
        resources:
          requests:
            cpu: 3
            memory: 5Gi

实际只运行了5个，实际应该运行至少6个Pod

$  kgp -A -o wide|grep demo-nginx
demo               demo-nginx-69db9d45df-4j2rh                                  1/1     Running                  0                  3h20m   172.30.64.199   ip-172-30-64-34.ap-northeast-1.compute.internal    <none>           <none>
demo               demo-nginx-69db9d45df-4jc5h                                  0/1     Pending                  0                  3h20m   <none>          <none>                                             <none>           <none>
demo               demo-nginx-69db9d45df-6p4jz                                  0/1     Pending                  0                  3h20m   <none>          <none>                                             <none>           <none>
demo               demo-nginx-69db9d45df-7fdn2                                  1/1     Running                  0                  3h20m   172.30.64.111   ip-172-30-64-34.ap-northeast-1.compute.internal    <none>           <none>
demo               demo-nginx-69db9d45df-b75mz                                  1/1     Running                  0                  3h20m   172.30.64.78    ip-172-30-64-34.ap-northeast-1.compute.internal    <none>           <none>
demo               demo-nginx-69db9d45df-vsp6g                                  1/1     Running                  0                  3h20m   172.30.64.97    ip-172-30-64-34.ap-northeast-1.compute.internal    <none>           <none>
demo               demo-nginx-69db9d45df-xxrsb                                  1/1     Running                  0                  3h20m   172.30.64.10    ip-172-30-64-34.ap-northeast-1.compute.internal    <none>           <none>
demo               demo-nginx-69db9d45df-zgkjr                                  0/1     Pending                  0                  8m56s   <none>          <none>                                             <none>           <none>

predicate配置

$ k get cm dynamic-scheduler-policy -n crane-system -o yaml 
apiVersion: v1
data:
  policy.yaml: |
    apiVersion: scheduler.policy.crane.io/v1alpha1
    kind: DynamicSchedulerPolicy
    spec:
      syncPolicy:
        ##cpu usage
        - name: cpu_usage_avg_5m
          period: 3m
        - name: cpu_usage_max_avg_1h
          period: 15m
        - name: cpu_usage_max_avg_1d
          period: 3h
        ##memory usage
        - name: mem_usage_avg_5m
          period: 3m
        - name: mem_usage_max_avg_1h
          period: 15m
        - name: mem_usage_max_avg_1d
          period: 3h

      predicate:
        ##cpu usage
        - name: cpu_usage_avg_5m
          maxLimitPecent: 0.90
        - name: cpu_usage_max_avg_1h
          maxLimitPecent: 0.95
        ##memory usage
        - name: mem_usage_avg_5m
          maxLimitPecent: 0.90
        - name: mem_usage_max_avg_1h
          maxLimitPecent: 0.95

      priority:
        ###score = sum(() * weight) / len,  0 <= score <= 10
        ##cpu usage
        - name: cpu_usage_avg_5m
          weight: 0.2
        - name: cpu_usage_max_avg_1h
          weight: 0.3
        - name: cpu_usage_max_avg_1d
          weight: 0.5
        ##memory usage
        - name: mem_usage_avg_5m
          weight: 0.2
        - name: mem_usage_max_avg_1h
          weight: 0.3
        - name: mem_usage_max_avg_1d
          weight: 0.5

      hotValue:
        - timeRange: 5m
          count: 20
        - timeRange: 1m
          count: 10

crane-scheduler是根据节点实际负载调度Pod，为什么节点内存负载是0.21，CPU负载是0.63，且未触发predicate指标阀值，实际只运行了5个pod，按照节点剩余25GB((1-0.21）32)内存、5核CPU((1-0.63）16)资源计算，至少可运行6个以上Pod

gocrane / crane-scheduler

没有根据节点实际负载调度Pod #51