Kube Scheduler

通过监控 Kube Scheduler 指标,帮助配置和优化Kube Scheduler，可以提高集群的资源利用率和应用程序的性能

配置¶

前置条件¶

已安装 DataKit

采集器配置¶

在 datakit.yaml 中配置 ConfigMap 资源来收集 Kube Scheduler 的指标数据

apiVersion: v1
kind: ConfigMap
metadata:
  name: datakit-conf
  namespace: datakit
data:
   kubernetesprometheus.conf: |-  
     [inputs.kubernetesprometheus]
        [[inputs.kubernetesprometheus.instances]]
          role       = "pod"
          namespaces = ["kube-system"]
          selector   = "component=kube_scheduler,tier=control-plane"      
          scrape   = "true"
          scheme   = "https"
          port     = "10259"
          path     = "/metrics"
          interval = "30s"

         [inputs.kubernetesprometheus.instances.custom]
           measurement        = "kube_scheduler"
           job_as_measurement = false
           [inputs.kubernetesprometheus.instances.custom.tags]
             cluster_name_k8s = "k8s-test"           
             node_name        = "__kubernetes_pod_node_name"
             instance         = "__kubernetes_mate_instance"

         [inputs.kubernetesprometheus.instances.auth]
           bearer_token_file = "/var/run/secrets/kubernetes.io/serviceaccount/token"
           [inputs.kubernetesprometheus.instances.auth.tls_config]
             insecure_skip_verify = true
             ca_certs = []
             cert     = ""
             cert_key = ""

重启 datakit

kubectl rollout restart ds  datakit -n datakit

指标¶

kube scheduler 指标集¶

Kube Scheduler 指标位于 kube-scheduler 指标集下，这里介绍 Kube Scheduler 指标相关说明

Metrics	描述	单位
`scheduler_scheduler_cache_size`	`调度器缓存中节点、Pod和AssumedPod（假定要调度的Pod）的数量`	count
`scheduler_pending_pods`	`Pending Pod的数量`	count
`process_cpu_seconds_total`	`kube-scheduler 进程的 CPU 使用总`	s
`rest_client_request_duration_seconds_bucket`	`从方法（Verb）和URL维度分析HTTP请求时延`	ms
`rest_client_requests_total`	`从状态值（Status Code）、方法（Method）和主机（Host）维度分析HTTP请求数`	count
`rest_client_requests_total`	`kube-scheduler 向 kube-apiserver 发起的 HTTP 请求总数`	count