Azure Monitor 中的默认 Prometheus 指标配置

本文列出了将 Prometheus 指标配置为从任何 Azure Kubernetes 服务 (AKS) 群集中抓取时的默认目标、仪表板和记录规则。

最小引入配置文件

Minimal ingestion profile 是一种有助于减少指标引入量的设置，因为只有默认仪表板使用的指标，才会收集默认记录规则和默认警报。 对于基于加载项的集合，Minimal ingestion profile 设置默认处于启用状态。 可以修改集合以启用收集更多指标，如下所示。

抓取频率

所有默认目标和抓取的默认抓取频率为 30 秒。

默认抓取的目标

以下目标默认处于“已启用/开启”状态，这意味着你不必提供任何抓取作业配置来抓取这些目标，因为指标加载项默认会自动抓取这些目标。

• cadvisor (job=cadvisor)
• nodeexporter (job=node)
• kubelet (job=kubelet)
• kube-state-metrics (job=kube-state-metrics)
• networkobservabilityRetina (job=networkobservabilityRetina)

启用控制平面指标（预览版）功能时，以下目标将启用/打开。 可以使用控制平面指标最大程度地提高整体可观测性，并保持 AKS 群集的卓越运营。

• controlplane-apiserver (job=controlplane-apiserver)
• controlplane-etcd (job=controlplane-etcd)

启用容器网络可观测性（这是高级容器网络服务套件的一项功能，它与所有 Linux 工作负载兼容，并与 Hubble 无缝集成，对于基于 Cilium 和非 Cilium 的数据平面均适用）时，以下目标将启用/打开。 这样可以灵活地满足容器网络需求。

• networkobservabilityHubble (job=networkobservabilityHubble)
• networkobservabilityCilium (job=networkobservabilityCilium)

从默认目标收集的指标

默认情况下，将从每个默认目标收集以下指标。 所有其他指标都将通过重新标记规则删除。

cadvisor (job=cadvisor)

• container_spec_cpu_period
• container_spec_cpu_quota
• container_cpu_usage_seconds_total
• container_memory_rss
• container_network_receive_bytes_total
• container_network_transmit_bytes_total
• container_network_receive_packets_total
• container_network_transmit_packets_total
• container_network_receive_packets_dropped_total
• container_network_transmit_packets_dropped_total
• container_fs_reads_total
• container_fs_writes_total
• container_fs_reads_bytes_total
• container_fs_writes_bytes_total
• container_memory_working_set_bytes
• container_memory_cache
• container_memory_swap
• container_cpu_cfs_throttled_periods_total
• container_cpu_cfs_periods_total
• container_memory_usage_bytes
• kubernetes_build_info"

kubelet (job=kubelet)

• kubelet_volume_stats_used_bytes
• kubelet_node_name
• kubelet_running_pods
• kubelet_running_pod_count
• kubelet_running_containers
• kubelet_running_container_count
• volume_manager_total_volumes
• kubelet_node_config_error
• kubelet_runtime_operations_total
• kubelet_runtime_operations_errors_total
• kubelet_runtime_operations_duration_seconds kubelet_runtime_operations_duration_seconds_bucket kubelet_runtime_operations_duration_seconds_sum kubelet_runtime_operations_duration_seconds_count
• kubelet_pod_start_duration_seconds kubelet_pod_start_duration_seconds_bucket kubelet_pod_start_duration_seconds_sum kubelet_pod_start_duration_seconds_count
• kubelet_pod_worker_duration_seconds kubelet_pod_worker_duration_seconds_bucket kubelet_pod_worker_duration_seconds_sum kubelet_pod_worker_duration_seconds_count
• storage_operation_duration_seconds storage_operation_duration_seconds_bucket storage_operation_duration_seconds_sum storage_operation_duration_seconds_count
• storage_operation_errors_total
• kubelet_cgroup_manager_duration_seconds kubelet_cgroup_manager_duration_seconds_bucket kubelet_cgroup_manager_duration_seconds_sum kubelet_cgroup_manager_duration_seconds_count
• kubelet_pleg_relist_duration_seconds kubelet_pleg_relist_duration_seconds_bucket kubelet_pleg_relist_duration_sum kubelet_pleg_relist_duration_seconds_count
• kubelet_pleg_relist_interval_seconds kubelet_pleg_relist_interval_seconds_bucket kubelet_pleg_relist_interval_seconds_sum kubelet_pleg_relist_interval_seconds_count
• rest_client_requests_total
• rest_client_request_duration_seconds rest_client_request_duration_seconds_bucket rest_client_request_duration_seconds_sum rest_client_request_duration_seconds_count
• process_resident_memory_bytes
• process_cpu_seconds_total
• go_goroutines
• kubelet_volume_stats_capacity_bytes
• kubelet_volume_stats_available_bytes
• kubelet_volume_stats_inodes_used
• kubelet_volume_stats_inodes
• kubernetes_build_info"

nodexporter (job=node)

• node_cpu_seconds_total
• node_memory_MemAvailable_bytes
• node_memory_Buffers_bytes
• node_memory_Cached_bytes
• node_memory_MemFree_bytes
• node_memory_Slab_bytes
• node_memory_MemTotal_bytes
• node_netstat_Tcp_RetransSegs
• node_netstat_Tcp_OutSegs
• node_netstat_TcpExt_TCPSynRetrans
• node_load1``node_load5
• node_load15
• node_disk_read_bytes_total
• node_disk_written_bytes_total
• node_disk_io_time_seconds_total
• node_filesystem_size_bytes
• node_filesystem_avail_bytes
• node_filesystem_readonly
• node_network_receive_bytes_total
• node_network_transmit_bytes_total
• node_vmstat_pgmajfault
• node_network_receive_drop_total
• node_network_transmit_drop_total
• node_disk_io_time_weighted_seconds_total
• node_exporter_build_info
• node_time_seconds
• node_uname_info"

kube-state-metrics (job=kube-state-metrics)

• kube_job_status_succeeded
• kube_job_spec_completions
• kube_daemonset_status_desired_number_scheduled
• kube_daemonset_status_number_ready
• kube_deployment_status_replicas_ready
• kube_pod_container_status_last_terminated_reason
• kube_pod_container_status_waiting_reason
• kube_pod_container_status_restarts_total
• kube_node_status_allocatable
• kube_pod_owner
• kube_pod_container_resource_requests
• kube_pod_status_phase
• kube_pod_container_resource_limits
• kube_replicaset_owner
• kube_resourcequota
• kube_namespace_status_phase
• kube_node_status_capacity
• kube_node_info
• kube_pod_info
• kube_deployment_spec_replicas
• kube_deployment_status_replicas_available
• kube_deployment_status_replicas_updated
• kube_statefulset_status_replicas_ready
• kube_statefulset_status_replicas
• kube_statefulset_status_replicas_updated
• kube_job_status_start_time
• kube_job_status_active
• kube_job_failed
• kube_horizontalpodautoscaler_status_desired_replicas
• kube_horizontalpodautoscaler_status_current_replicas
• kube_horizontalpodautoscaler_spec_min_replicas
• kube_horizontalpodautoscaler_spec_max_replicas
• kubernetes_build_info
• kube_node_status_condition
• kube_node_spec_taint
• kube_pod_container_info
• kube_resource_labels (ex - kube_pod_labels, kube_deployment_labels)
• kube_resource_annotations (ex - kube_pod_annotations, kube_deployment_annotations)

controlplane-apiserver (job=controlplane-apiserver)

• apiserver_request_total
• apiserver_cache_list_fetched_objects_total
• apiserver_cache_list_returned_objects_total
• apiserver_flowcontrol_demand_seats_average
• apiserver_flowcontrol_current_limit_seats
• apiserver_request_sli_duration_seconds_bucket
• apiserver_request_sli_duration_seconds_count
• apiserver_request_sli_duration_seconds_sum
• process_start_time_seconds
• apiserver_request_duration_seconds_bucket
• apiserver_request_duration_seconds_count
• apiserver_request_duration_seconds_sum
• apiserver_storage_list_fetched_objects_total
• apiserver_storage_list_returned_objects_total
• apiserver_current_inflight_requests

controlplane-etcd (job=controlplane-etcd)

• etcd_server_has_leader
• rest_client_requests_total
• etcd_mvcc_db_total_size_in_bytes
• etcd_mvcc_db_total_size_in_use_in_bytes
• etcd_server_slow_read_indexes_total
• etcd_server_slow_apply_total
• etcd_network_client_grpc_sent_bytes_total
• etcd_server_heartbeat_send_failures_total

networkobservabilityHubble (job=networkobservabilityHubble) 和 networkobservabilityCilium (job=networkobservabilityCilium)

默认抓取的 Windows 目标

以下 Windows 目标配置为抓取，但默认情况下抓取未启用（已禁用/关闭），这意味着你不必提供任何抓取作业配置来抓取这些目标，但它们默认处于禁用/关闭状态，你需要使用 default-scrape-settings-enabled 部分下的 ama-metrics-settings-configmap 为这些目标开启/启用抓取。

可以针对 Windows 运行两个默认作业，这些作业将抓取特定于 Windows 的仪表板所需的指标。

• windows-exporter (job=windows-exporter)
• kube-proxy-windows (job=kube-proxy-windows)

为 Windows 抓取的指标

当启用 windows-exporter 和 kube-proxy-windows 时，将收集以下指标。

windows-exporter (job=windows-exporter)

• windows_system_system_up_time
• windows_cpu_time_total
• windows_memory_available_bytes
• windows_os_visible_memory_bytes
• windows_memory_cache_bytes
• windows_memory_modified_page_list_bytes
• windows_memory_standby_cache_core_bytes
• windows_memory_standby_cache_normal_priority_bytes
• windows_memory_standby_cache_reserve_bytes
• windows_memory_swap_page_operations_total
• windows_logical_disk_read_seconds_total
• windows_logical_disk_write_seconds_total
• windows_logical_disk_size_bytes
• windows_logical_disk_free_bytes
• windows_net_bytes_total
• windows_net_packets_received_discarded_total
• windows_net_packets_outbound_discarded_total
• windows_container_available
• windows_container_cpu_usage_seconds_total
• windows_container_memory_usage_commit_bytes
• windows_container_memory_usage_private_working_set_bytes
• windows_container_network_receive_bytes_total
• windows_container_network_transmit_bytes_total

kube-proxy-windows (job=kube-proxy-windows)

• kubeproxy_sync_proxy_rules_duration_seconds
• kubeproxy_sync_proxy_rules_duration_seconds_bucket
• kubeproxy_sync_proxy_rules_duration_seconds_sum
• kubeproxy_sync_proxy_rules_duration_seconds_count
• rest_client_requests_total
• rest_client_request_duration_seconds
• rest_client_request_duration_seconds_bucket
• rest_client_request_duration_seconds_sum
• rest_client_request_duration_seconds_count
• process_resident_memory_bytes
• process_cpu_seconds_total
• go_goroutines

仪表板

当你将 Azure Monitor 工作区链接到 Azure 托管 Grafana 实例时，适用于 Prometheus 的 Azure Monitor 托管服务将自动预配和配置以下默认仪表板。 这些仪表板的源代码可在此 GitHub 存储库中找到。 以下仪表板将在 Grafana 中 Managed Prometheus 文件夹下的指定 Azure Grafana 实例中预配。 这些是标准开放源代码社区仪表板，可通过 Prometheus 和 Grafana 监视 Kubernetes 群集。

• Kubernetes / Compute Resources / Cluster
• Kubernetes / Compute Resources / Namespace (Pods)
• Kubernetes / Compute Resources / Node (Pods)
• Kubernetes / Compute Resources / Pod
• Kubernetes / Compute Resources / Namespace (Workloads)
• Kubernetes / Compute Resources / Workload
• Kubernetes / Kubelet
• Node Exporter / USE Method / Node
• Node Exporter / Nodes
• Kubernetes / Compute Resources / Cluster (Windows)
• Kubernetes / Compute Resources / Namespace (Windows)
• Kubernetes / Compute Resources / Pod (Windows)
• Kubernetes / USE Method / Cluster (Windows)
• Kubernetes / USE Method / Node (Windows)

记录规则

当你配置要从 Azure Kubernetes 服务 (AKS) 群集中抓取的 Prometheus 指标时，适用于 Prometheus 的 Azure Monitor 托管服务会自动配置以下默认记录规则。 这些记录规则的源代码可在此 GitHub 存储库中找到。 这些是上述仪表板中使用的标准开放源代码记录规则。

• cluster:node_cpu:ratio_rate5m
• namespace_cpu:kube_pod_container_resource_requests:sum
• namespace_cpu:kube_pod_container_resource_limits:sum
• :node_memory_MemAvailable_bytes:sum
• namespace_memory:kube_pod_container_resource_requests:sum
• namespace_memory:kube_pod_container_resource_limits:sum
• namespace_workload_pod:kube_pod_owner:relabel
• node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate
• cluster:namespace:pod_cpu:active:kube_pod_container_resource_requests
• cluster:namespace:pod_cpu:active:kube_pod_container_resource_limits
• cluster:namespace:pod_memory:active:kube_pod_container_resource_requests
• cluster:namespace:pod_memory:active:kube_pod_container_resource_limits
• node_namespace_pod_container:container_memory_working_set_bytes
• node_namespace_pod_container:container_memory_rss
• node_namespace_pod_container:container_memory_cache
• node_namespace_pod_container:container_memory_swap
• instance:node_cpu_utilisation:rate5m
• instance:node_load1_per_cpu:ratio
• instance:node_memory_utilisation:ratio
• instance:node_vmstat_pgmajfault:rate5m
• instance:node_network_receive_bytes_excluding_lo:rate5m
• instance:node_network_transmit_bytes_excluding_lo:rate5m
• instance:node_network_receive_drop_excluding_lo:rate5m
• instance:node_network_transmit_drop_excluding_lo:rate5m
• instance_device:node_disk_io_time_seconds:rate5m
• instance_device:node_disk_io_time_weighted_seconds:rate5m
• instance:node_num_cpu:sum
• node:windows_node:sum
• node:windows_node_num_cpu:sum
• :windows_node_cpu_utilisation:avg5m
• node:windows_node_cpu_utilisation:avg5m
• :windows_node_memory_utilisation:
• :windows_node_memory_MemFreeCached_bytes:sum
• node:windows_node_memory_totalCached_bytes:sum
• :windows_node_memory_MemTotal_bytes:sum
• node:windows_node_memory_bytes_available:sum
• node:windows_node_memory_bytes_total:sum
• node:windows_node_memory_utilisation:ratio
• node:windows_node_memory_utilisation:
• node:windows_node_memory_swap_io_pages:irate
• :windows_node_disk_utilisation:avg_irate
• node:windows_node_disk_utilisation:avg_irate
• node:windows_node_filesystem_usage:
• node:windows_node_filesystem_avail:
• :windows_node_net_utilisation:sum_irate
• node:windows_node_net_utilisation:sum_irate
• :windows_node_net_saturation:sum_irate
• node:windows_node_net_saturation:sum_irate
• windows_pod_container_available
• windows_container_total_runtime
• windows_container_memory_usage
• windows_container_private_working_set_usage
• windows_container_network_received_bytes_total
• windows_container_network_transmitted_bytes_total
• kube_pod_windows_container_resource_memory_request
• kube_pod_windows_container_resource_memory_limit
• kube_pod_windows_container_resource_cpu_cores_request
• kube_pod_windows_container_resource_cpu_cores_limit
• namespace_pod_container:windows_container_cpu_usage_seconds_total:sum_rate

Prometheus 可视化效果记录规则

当使用基于 Prometheus 的容器见解时，将部署更多记录规则以支持 Prometheus 可视化效果。

• ux:cluster_pod_phase_count:sum
• ux:node_cpu_usage:sum_irate
• ux:node_memory_usage:sum
• ux:controller_pod_phase_count:sum
• ux:controller_container_count:sum
• ux:controller_workingset_memory:sum
• ux:controller_cpu_usage:sum_irate
• ux:controller_rss_memory:sum
• ux:controller_resource_limit:sum
• ux:controller_container_restarts:max
• ux:pod_container_count:sum
• ux:pod_cpu_usage:sum_irate
• ux:pod_workingset_memory:sum
• ux:pod_rss_memory:sum
• ux:pod_resource_limit:sum
• ux:pod_container_restarts:max
• ux:node_network_receive_drop_total:sum_irate
• ux:node_network_transmit_drop_total:sum_irate

Windows 支持需要以下记录规则。 这些规则与上述规则一起部署，但默认情况下并未启用。

• ux:node_cpu_usage_windows:sum_irate
• ux:node_memory_usage_windows:sum
• ux:controller_cpu_usage_windows:sum_irate
• ux:controller_workingset_memory_windows:sum
• ux:pod_cpu_usage_windows:sum_irate
• ux:pod_workingset_memory_windows:sum

后续步骤

自定义 Prometheus 指标抓取。