dashboard

监控项：

• 各个k8s集群所有node的状态，包括node NotReady/Unschedulabel，设置dashboard refresh every 30s，可实时看到node的状态，同时可设置grafana告警

• node的kernel/docker/kubelet版本差异统计，并支持过滤

  
  k8s-node-status.png
  k8s-node-status2.png

TechStack

promtheus+grafana

PromQL

datasource接入prometheus，PromQL语句对应上图各项title

• NotReady

sum(kube_node_status_condition{condition="Ready",status="unknown",node=~"$node"})

• Unschedulable

sum(kube_node_spec_unschedulable{node=~"$node"})

• Kubelet Version Status

sum(kube_node_info{node=~"$node"}) by (kubelet_version)

• Docker Version Status

sum(kube_node_info{node=~"$node"}) by (container_runtime_version)

• Kernel Version Status

sum(kube_node_info{node=~"$node"}) by (kernel_version)

• K8S Node Status

(kube_node_status_condition{condition="Ready",status="unknown",node=~"$node"}) * on(node) group_right() (sum(kube_node_labels) by (label_FOO_com_role,node)*on(node) group_right(label_cloud_ctrip_com_role) sum(kube_node_info{container_runtime_version=~"$docker_version",kernel_version=~"$kernel_version",kubelet_version=~"$kubelet_version"}) by (container_runtime_version,kernel_version,kubelet_version,node))