Container Service for Kubernetes:複数クラスターのアラート管理

有効にするルールグループの選択

テンプレートには 11 のルールグループが含まれています。ご利用のクラスター設定に関連するグループのみを有効にしてください：

アラートルールの適用

1. 有効にしたいルールグループの rules.enable を enable に設定します。以下の例では、error-events が有効になっています。

2. ステップ 2 の contactGroups ブロックを追加します。

3. ファイルを ackalertrule.yaml として保存し、適用します：

   kubectl apply -f ackalertrule.yaml

error-events を有効にした完全な例を以下に示します：

apiVersion: alert.alibabacloud.com/v1beta1
kind: AckAlertRule
metadata:
  name: default
  namespace: kube-system
spec:
  groups:
  - name: error-events
    rules:
    - enable: enable
      contactGroups:
      - arms_contact_group_id: "1****"
        cms_contact_group_name: ack_Default Contact Group
        id: "3***"
      expression: sls.app.ack.error
      name: error-event
      notification:
        message: kubernetes cluster error event.
      type: event
  - name: warn-events
    rules:
    - enable: disable
      expression: sls.app.ack.warn
      name: warn-event
      notification:
        message: kubernetes cluster warn event.
      type: event
  - name: cluster-core-error
    rules:
    - enable: disable
      expression: prom.apiserver.notHealthy.down
      name: apiserver-unhealthy
      notification:
        message: "Cluster APIServer not healthy. \nPromQL: ((sum(up{job=\"apiserver\"})
          <= 0) or (absent(sum(up{job=\"apiserver\"})))) > 0"
      type: metric-prometheus
    - enable: disable
      expression: prom.etcd.notHealthy.down
      name: etcd-unhealthy
      notification:
        message: "Cluster ETCD not healthy. \nPromQL: ((sum(up{job=\"etcd\"}) <= 0)
          or (absent(sum(up{job=\"etcd\"})))) > 0"
      type: metric-prometheus
    - enable: disable
      expression: prom.scheduler.notHealthy.down
      name: scheduler-unhealthy
      notification:
        message: "Cluster Scheduler not healthy. \nPromQL: ((sum(up{job=\"ack-scheduler\"})
          <= 0) or (absent(sum(up{job=\"ack-scheduler\"})))) > 0"
      type: metric-prometheus
    - enable: disable
      expression: prom.kcm.notHealthy.down
      name: kcm-unhealthy
      notification:
        message: "Custer kube-controller-manager not healthy. \nPromQL: ((sum(up{job=\"ack-kube-controller-manager\"})
          <= 0) or (absent(sum(up{job=\"ack-kube-controller-manager\"})))) > 0"
      type: metric-prometheus
    - enable: disable
      expression: prom.ccm.notHealthy.down
      name: ccm-unhealthy
      notification:
        message: "Cluster cloud-controller-manager not healthy. \nPromQL: ((sum(up{job=\"ack-cloud-controller-manager\"})
          <= 0) or (absent(sum(up{job=\"ack-cloud-controller-manager\"})))) > 0"
      type: metric-prometheus
    - enable: disable
      expression: prom.coredns.notHealthy.requestdown
      name: coredns-unhealthy-requestdown
      notification:
        message: "Cluster CoreDNS not healthy, continuously request down. \nPromQL:
          (sum(rate(coredns_dns_request_count_total{}[1m]))by(server,zone)<=0) or
          (sum(rate(coredns_dns_requests_total{}[1m]))by(server,zone)<=0)"
      type: metric-prometheus
    - enable: disable
      expression: prom.coredns.notHealthy.panic
      name: coredns-unhealthy-panic
      notification:
        message: "Cluster CoreDNS not healthy, continuously panic. \nPromQL: sum(rate(coredns_panic_count_total{}[3m]))
          > 0"
      type: metric-prometheus
    - enable: disable
      expression: prom.ingress.request.errorRateHigh
      name: ingress-err-request
      notification:
        message: Cluster Ingress Controller request error rate high (default error
          rate is 85%).
      type: metric-prometheus
    - enable: disable
      expression: prom.ingress.ssl.expire
      name: ingress-ssl-expire
      notification:
        message: "Cluster Ingress Controller SSL will expire in a few days (default
          14 days). \nPromQL: ((nginx_ingress_controller_ssl_expire_time_seconds -
          time()) / 24 / 3600) < 14"
      type: metric-prometheus
  - name: cluster-error
    rules:
    - enable: disable
      expression: sls.app.ack.docker.hang
      name: docker-hang
      notification:
        message: kubernetes node docker hang.
      type: event
    - enable: disable
      expression: sls.app.ack.eviction
      name: eviction-event
      notification:
        message: kubernetes eviction event.
      type: event
    - enable: disable
      expression: sls.app.ack.gpu.xid_error
      name: gpu-xid-error
      notification:
        message: kubernetes gpu xid error event.
      type: event
    - enable: disable
      expression: sls.app.ack.image.pull_back_off
      name: image-pull-back-off
      notification:
        message: kubernetes image pull back off event.
      type: event
    - enable: disable
      expression: sls.app.ack.node.down
      name: node-down
      notification:
        message: kubernetes node down event.
      type: event
    - enable: disable
      expression: sls.app.ack.node.restart
      name: node-restart
      notification:
        message: kubernetes node restart event.
      type: event
    - enable: disable
      expression: sls.app.ack.ntp.down
      name: node-ntp-down
      notification:
        message: kubernetes node ntp down.
      type: event
    - enable: disable
      expression: sls.app.ack.node.pleg_error
      name: node-pleg-error
      notification:
        message: kubernetes node pleg error event.
      type: event
    - enable: disable
      expression: sls.app.ack.ps.hang
      name: ps-hang
      notification:
        message: kubernetes ps hang event.
      type: event
    - enable: disable
      expression: sls.app.ack.node.fd_pressure
      name: node-fd-pressure
      notification:
        message: kubernetes node fd pressure event.
      type: event
    - enable: disable
      expression: sls.app.ack.node.pid_pressure
      name: node-pid-pressure
      notification:
        message: kubernetes node pid pressure event.
      type: event
    - enable: disable
      expression: sls.app.ack.ccm.del_node_failed
      name: node-del-err
      notification:
        message: kubernetes delete node failed.
      type: event
    - enable: disable
      expression: sls.app.ack.ccm.add_node_failed
      name: node-add-err
      notification:
        message: kubernetes add node failed.
      type: event
    - enable: disable
      expression: sls.app.ack.nlc.run_command_fail
      name: nlc-run-cmd-err
      notification:
        message: kubernetes node pool nlc run command failed.
      type: event
    - enable: disable
      expression: sls.app.ack.nlc.empty_task_cmd
      name: nlc-empty-cmd
      notification:
        message: kubernetes node pool nlc delete node failed.
      type: event
    - enable: disable
      expression: sls.app.ack.nlc.url_mode_unimpl
      name: nlc-url-m-unimp
      notification:
        message: kubernetes nodde pool nlc delete node failed.
      type: event
    - enable: disable
      expression: sls.app.ack.nlc.op_not_found
      name: nlc-opt-no-found
      notification:
        message: kubernetes node pool nlc delete node failed.
      type: event
    - enable: disable
      expression: sls.app.ack.nlc.destroy_node_fail
      name: nlc-des-node-err
      notification:
        message: kubernetes node pool nlc destory node failed.
      type: event
    - enable: disable
      expression: sls.app.ack.nlc.drain_node_fail
      name: nlc-drain-node-err
      notification:
        message: kubernetes node pool nlc drain node failed.
      type: event
    - enable: disable
      expression: sls.app.ack.nlc.restart_ecs_wait_fail
      name: nlc-restart-ecs-wait
      notification:
        message: kubernetes node pool nlc restart ecs wait timeout.
      type: event
    - enable: disable
      expression: sls.app.ack.nlc.restart_ecs_fail
      name: nlc-restart-ecs-err
      notification:
        message: kubernetes node pool nlc restart ecs failed.
      type: event
    - enable: disable
      expression: sls.app.ack.nlc.reset_ecs_fail
      name: nlc-reset-ecs-err
      notification:
        message: kubernetes node pool nlc reset ecs failed.
      type: event
    - enable: disable
      expression: sls.app.ack.nlc.repair_fail
      name: nlc-sel-repair-err
      notification:
        message: kubernetes node pool nlc self repair failed.
      type: event
  - name: res-exceptions
    rules:
    - enable: disable
      expression: cms.host.cpu.utilization
      name: node_cpu_util_high
      notification:
        message: kubernetes cluster node cpu utilization too high.
      thresholds:
      - key: CMS_ESCALATIONS_CRITICAL_Threshold
        unit: percent
        value: "85"
      type: metric-cms
    - enable: disable
      expression: cms.host.memory.utilization
      name: node_mem_util_high
      notification:
        message: kubernetes cluster node memory utilization too high.
      thresholds:
      - key: CMS_ESCALATIONS_CRITICAL_Threshold
        unit: percent
        value: "85"
      type: metric-cms
    - enable: disable
      expression: cms.host.disk.utilization
      name: node_disk_util_high
      notification:
        message: kubernetes cluster node disk utilization too high.
      thresholds:
      - key: CMS_ESCALATIONS_CRITICAL_Threshold
        unit: percent
        value: "85"
      type: metric-cms
    - enable: disable
      expression: cms.host.public.network.utilization
      name: node_public_net_util_high
      notification:
        message: kubernetes cluster node public network utilization too high.
      thresholds:
      - key: CMS_ESCALATIONS_CRITICAL_Threshold
        unit: percent
        value: "85"
      type: metric-cms
    - enable: disable
      expression: cms.host.fs.inode.utilization
      name: node_fs_inode_util_high
      notification:
        message: kubernetes cluster node file system inode utilization too high.
      thresholds:
      - key: CMS_ESCALATIONS_CRITICAL_Threshold
        unit: percent
        value: "85"
      type: metric-cms
    - enable: disable
      expression: cms.slb.qps.utilization
      name: slb_qps_util_high
      notification:
        message: kubernetes cluster slb qps utilization too high.
      thresholds:
      - key: CMS_ESCALATIONS_CRITICAL_Threshold
        unit: percent
        value: "85"
      type: metric-cms
    - enable: disable
      expression: cms.slb.traffic.tx.utilization
      name: slb_traff_tx_util_high
      notification:
        message: kubernetes cluster slb traffic utilization too high.
      thresholds:
      - key: CMS_ESCALATIONS_CRITICAL_Threshold
        unit: percent
        value: "85"
      type: metric-cms
    - enable: disable
      expression: cms.slb.max.connection.utilization
      name: slb_max_con_util_high
      notification:
        message: kubernetes cluster max connection utilization too high.
      thresholds:
      - key: CMS_ESCALATIONS_CRITICAL_Threshold
        unit: percent
        value: "85"
      type: metric-cms
    - enable: disable
      expression: cms.slb.drop.connection
      name: slb_drop_con_high
      notification:
        message: kubernetes cluster drop connection count per second too high.
      thresholds:
      - key: CMS_ESCALATIONS_CRITICAL_Threshold
        unit: count
        value: "1"
      type: metric-cms
    - enable: disable
      expression: sls.app.ack.node.disk_pressure
      name: node-disk-pressure
      notification:
        message: kubernetes node disk pressure event.
      type: event
    - enable: disable
      expression: sls.app.ack.resource.insufficient
      name: node-res-insufficient
      notification:
        message: kubernetes node resource insufficient.
      type: event
    - enable: disable
      expression: sls.app.ack.ip.not_enough
      name: node-ip-pressure
      notification:
        message: kubernetes ip not enough event.
      type: event
    - enable: disable
      expression: sls.app.ack.csi.no_enough_disk_space
      name: disk_space_press
      notification:
        message: kubernetes csi not enough disk space.
      type: event
  - name: cluster-scale
    rules:
    - enable: disable
      expression: sls.app.ack.autoscaler.scaleup_group
      name: autoscaler-scaleup
      notification:
        message: kubernetes autoscaler scale up.
      type: event
    - enable: disable
      expression: sls.app.ack.autoscaler.scaledown
      name: autoscaler-scaledown
      notification:
        message: kubernetes autoscaler scale down.
      type: event
    - enable: disable
      expression: sls.app.ack.autoscaler.scaleup_timeout
      name: autoscaler-scaleup-timeout
      notification:
        message: kubernetes autoscaler scale up timeout.
      type: event
    - enable: disable
      expression: sls.app.ack.autoscaler.scaledown_empty
      name: autoscaler-scaledown-empty
      notification:
        message: kubernetes autoscaler scale down empty node.
      type: event
    - enable: disable
      expression: sls.app.ack.autoscaler.scaleup_group_failed
      name: autoscaler-up-group-failed
      notification:
        message: kubernetes autoscaler scale up failed.
      type: event
    - enable: disable
      expression: sls.app.ack.autoscaler.cluster_unhealthy
      name: autoscaler-cluster-unhealthy
      notification:
        message: kubernetes autoscaler error, cluster not healthy.
      type: event
    - enable: disable
      expression: sls.app.ack.autoscaler.delete_started_timeout
      name: autoscaler-del-started
      notification:
        message: kubernetes autoscaler delete node started long ago.
      type: event
    - enable: disable
      expression: sls.app.ack.autoscaler.delete_unregistered
      name: autoscaler-del-unregistered
      notification:
        message: kubernetes autoscaler delete unregistered node.
      type: event
    - enable: disable
      expression: sls.app.ack.autoscaler.scaledown_failed
      name: autoscaler-scale-down-failed
      notification:
        message: kubernetes autoscaler scale down failed.
      type: event
    - enable: disable
      expression: sls.app.ack.autoscaler.instance_expired
      name: autoscaler-instance-expired
      notification:
        message: kubernetes autoscaler scale down instance expired.
      type: event
  - name: workload-exceptions
    rules:
    - enable: disable
      expression: prom.job.failed
      name: job-failed
      notification:
        message: "Cluster Job failed. \nPromQL: kube_job_status_failed{job=\"_kube-state-metrics\"}
          > 0"
      type: metric-prometheus
    - enable: disable
      expression: prom.deployment.replicaError
      name: deployment-rep-err
      notification:
        message: "Cluster Deployment replication status error. \nPromQL: kube_deployment_spec_replicas{job=\"_kube-state-metrics\"}
          != kube_deployment_status_replicas_available{job=\"_kube-state-metrics\"}"
      type: metric-prometheus
    - enable: disable
      expression: prom.daemonset.scheduledError
      name: daemonset-status-err
      notification:
        message: "Cluster Daemonset pod status or scheduled error. \nPromQL: ((100
          - kube_daemonset_status_number_ready{} / kube_daemonset_status_desired_number_scheduled{}
          * 100) or (kube_daemonset_status_desired_number_scheduled{} - kube_daemonset_status_current_number_scheduled{}))
          > 0"
      type: metric-prometheus
    - enable: disable
      expression: prom.daemonset.misscheduled
      name: daemonset-misscheduled
      notification:
        message: "Cluster Daemonset misscheduled. \nPromQL: kube_daemonset_status_number_misscheduled{job=\"_kube-state-metrics\"}
          \ > 0"
      type: metric-prometheus
  - name: pod-exceptions
    rules:
    - enable: disable
      expression: sls.app.ack.pod.oom
      name: pod-oom
      notification:
        message: kubernetes pod oom event.
      type: event
    - enable: disable
      expression: sls.app.ack.pod.failed
      name: pod-failed
      notification:
        message: kubernetes pod start failed event.
      type: event
    - enable: disable
      expression: prom.pod.status.notHealthy
      name: pod-status-err
      notification:
        message: 'Pod status exception. \nPromQL: min_over_time(sum by (namespace,
          pod, phase) (kube_pod_status_phase{phase=~"Pending|Unknown|Failed", job="_kube-state-metrics"})[${mins}m:1m])
          > 0'
      type: metric-prometheus
    - enable: disable
      expression: prom.pod.status.crashLooping
      name: pod-crashloop
      notification:
        message: 'Pod status exception. \nPromQL: sum_over_time(increase(kube_pod_container_status_restarts_total{job="_kube-state-metrics"}[1m])[${mins}m:1m])
          > 3'
      type: metric-prometheus
  - name: cluster-storage-err
    rules:
    - enable: disable
      expression: sls.app.ack.csi.invalid_disk_size
      name: csi_invalid_size
      notification:
        message: kubernetes csi invalid disk size.
      type: event
    - enable: disable
      expression: sls.app.ack.csi.disk_not_portable
      name: csi_not_portable
      notification:
        message: kubernetes csi not protable.
      type: event
    - enable: disable
      expression: sls.app.ack.csi.deivce_busy
      name: csi_device_busy
      notification:
        message: kubernetes csi disk device busy.
      type: event
    - enable: disable
      expression: sls.app.ack.csi.no_ava_disk
      name: csi_no_ava_disk
      notification:
        message: kubernetes csi no available disk.
      type: event
    - enable: disable
      expression: sls.app.ack.csi.disk_iohang
      name: csi_disk_iohang
      notification:
        message: kubernetes csi ioHang.
      type: event
    - enable: disable
      expression: sls.app.ack.csi.latency_too_high
      name: csi_latency_high
      notification:
        message: kubernetes csi pvc latency load too high.
      type: event
    - enable: disable
      expression: prom.pv.failed
      name: pv-failed
      notification:
        message: 'Cluster PersistentVolume failed. \nPromQL: kube_persistentvolume_status_phase{phase=~"Failed|Pending",
          job="_kube-state-metrics"} > 0'
      type: metric-prometheus
  - name: cluster-network-err
    rules:
    - enable: disable
      expression: sls.app.ack.ccm.no_ava_slb
      name: slb-no-ava
      notification:
        message: kubernetes slb not available.
      type: event
    - enable: disable
      expression: sls.app.ack.ccm.sync_slb_failed
      name: slb-sync-err
      notification:
        message: kubernetes slb sync failed.
      type: event
    - enable: disable
      expression: sls.app.ack.ccm.del_slb_failed
      name: slb-del-err
      notification:
        message: kubernetes slb delete failed.
      type: event
    - enable: disable
      expression: sls.app.ack.ccm.create_route_failed
      name: route-create-err
      notification:
        message: kubernetes create route failed.
      type: event
    - enable: disable
      expression: sls.app.ack.ccm.sync_route_failed
      name: route-sync-err
      notification:
        message: kubernetes sync route failed.
      type: event
    - enable: disable
      expression: sls.app.ack.terway.invalid_resource
      name: terway-invalid-res
      notification:
        message: kubernetes terway have invalid resource.
      type: event
    - enable: disable
      expression: sls.app.ack.terway.alloc_ip_fail
      name: terway-alloc-ip-err
      notification:
        message: kubernetes terway allocate ip error.
      type: event
    - enable: disable
      expression: sls.app.ack.terway.parse_fail
      name: terway-parse-err
      notification:
        message: kubernetes terway parse k8s.aliyun.com/ingress-bandwidth annotation
          error.
      type: event
    - enable: disable
      expression: sls.app.ack.terway.allocate_failure
      name: terway-alloc-res-err
      notification:
        message: kubernetes parse resource error.
      type: event
    - enable: disable
      expression: sls.app.ack.terway.dispose_failure
      name: terway-dispose-err
      notification:
        message: kubernetes dispose resource error.
      type: event
    - enable: disable
      expression: sls.app.ack.terway.virtual_mode_change
      name: terway-virt-mod-err
      notification:
        message: kubernetes virtual mode changed.
      type: event
    - enable: disable
      expression: sls.app.ack.terway.config_check
      name: terway-ip-check
      notification:
        message: kubernetes terway execute pod ip config check.
      type: event
    - enable: disable
      expression: sls.app.ack.ingress.err_reload_nginx
      name: ingress-reload-err
      notification:
        message: kubernetes ingress reload config error.
      type: event
  - name: security-err
    rules:
    - enable: disable
      expression: sls.app.ack.si.config_audit_high_risk
      name: si-c-a-risk
      notification:
        message: kubernetes high risks have be found after running config audit.
      type: event
  ruleVersion: v1.0.9

アラートルールはフリートインスタンス上に作成されますが、ステップ 4 でディストリビューションルールを作成するまで、どのクラスターにも適用されません。

メソッド 1：ラベルによるディストリビューション

1. 関連付けられたクラスターの ID をクエリし、ターゲットにしたいクラスターにラベルを追加します：

   kubectl get managedclusters
   kubectl label managedclusters <clusterid> production=true

2. 次の内容で ackalertrule-app.yaml を作成します：

   apiVersion: core.oam.dev/v1beta1
   kind: Application
   metadata:
     name: alertrules
     namespace: kube-system
     annotations:
       app.oam.dev/publishVersion: version1
   spec:
     components:
       - name: alertrules
         type: ref-objects
         properties:
           objects:
             - resource: ackalertrules
               name: default
     policies:
       - type: topology
         name: prod-clusters
         properties:
           clusterSelector:
             production: "true"  # このラベルを持つクラスターを選択します

メソッド 2：クラスター ID によるディストリビューション

ターゲットクラスターの ID を指定して ackalertrule-app.yaml を作成します：

apiVersion: core.oam.dev/v1beta1
kind: Application
metadata:
  name: alertrules
  namespace: kube-system
  annotations:
    app.oam.dev/publishVersion: version1
spec:
  components:
    - name: alertrules
      type: ref-objects
      properties:
        objects:
          - resource: ackalertrules
            name: default
  policies:
    - type: topology
      name: prod-clusters
      properties:
        clusters: ["<clusterid1>", "<clusterid2>"]  # 実際のクラスター ID に置き換えます

ディストリビューションルールの適用と検証

1. ディストリビューションルールを適用します：

   kubectl apply -f ackalertrule-app.yaml

2. ディストリビューションのステータスを確認します：

   kubectl amc appstatus alertrules -n kube-system --tree --detail

   ディストリビューションが成功すると、各クラスターの出力に updated と表示されます：

   CLUSTER                  NAMESPACE       RESOURCE             STATUS    APPLY_TIME          DETAIL
   c565e4**** (cluster1)─── kube-system─── AckAlertRule/default updated   2022-**-** **:**:** Age: **
   cbaa12**** (cluster2)─── kube-system─── AckAlertRule/default updated   2022-**-** **:**:** Age: **

   クラスターのステータスが updated 以外の場合、そのクラスターがまだフリートインスタンスに関連付けられていること、および AckAlertRule リソースがフリートインスタンス上で正常に作成されたことを確認してください。アラート管理の詳細については、「アラート管理」をご参照ください。