【TKE】 GPU 监控 和 HPA 配置

IT知识
244
0
0
2024-08-02
【数据结构】算法效率的度量方法
上一篇
2023人工智能工程五大新方向
下一篇

原生节点 POD GPU 监控

如何查看原生节点 POD 基础监控？

前置条件：

TKE 提供 elastic-gpu-exporter 组件用于获取 GPU 相关监控指标，需要先安装该组件采集GPU 指标数据，参考基础监控指标采集：容器服务 GPU 监控指标获取-TKE 标准集群指南-文档中心-腾讯云

Pod 中使用了gpu 资源基础监控面板才会有 gpu的 tab，否则将不会显示， 如下图：

原生节点 POD GPU监控

超级节点 POD GPU 监控

如何查看超级节点 POD 基础监控？

超级节点pod 的 GPU 监控是默认显示的，可以直接看，如下图：

超级节点 POD GPU 监控

配置 GPU 指标的 HPA

使用场景

在业务使用 GPU 过程中，当业务处理量变大时，可能因 GPU 资源不足而导致业务出错，如下：

error:CUDA out of memory. Tried to allocate 112.00 MiB. GPU 0 has a total capacty of 14.75 GiB of which 46.81 MiB is free. Process 1075 has 14.70 GiB memory in use. Of the allocated memory 12.27 GiB is allocated by PyTorch, and 1.25 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

如何使用

可以根据 GPU 指标项通过 YAML 来配置 HPA，保障业务稳定运行。 譬如针对 GPU 资源的使用情况配置 HPA，相关配置示例如下：

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: example
  namespace: default
  labels:
    qcloud-app: example
spec:
  minReplicas: 1
  maxReplicas: 2
  metrics:
  - type: Pods	# 支持使用 Resource
    pods:
      metric:
        name: k8s_pod_gpu_memory_used_bytes # GPU 内存使用量
      target:
        averageValue: "10240" # 10240 MiB
        type: AverageValue
  scaleTargetRef:
    apiVersion: apps/v1beta2
    kind: Deployment
    name: nginx

更多指标参考HPA指标：容器服务 自动伸缩指标说明-TKE 标准集群指南-文档中心-腾讯云

配置 Grafana 面板

导入Grafana 面板数据

可以直接导入 TKE 整理的监控面板文件，如下：

{
  "annotations": {
    "list": [
      {
        "builtIn": 1,
        "datasource": "-- Grafana --",
        "enable": true,
        "hide": true,
        "iconColor": "rgba(0, 211, 255, 1)",
        "name": "Annotations & Alerts",
        "target": {
          "limit": 100,
          "matchAny": false,
          "tags": [],
          "type": "dashboard"
        },
        "type": "dashboard"
      }
    ]
  },
  "editable": true,
  "fiscalYearStartMonth": 0,
  "gnetId": null,
  "graphTooltip": 0,
  "id": 45,
  "iteration": 1663666644725,
  "links": [],
  "liveNow": false,
  "panels": [
    {
      "cacheTimeout": null,
      "datasource": "${datasource}",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [
            {
              "options": {
                "match": "null",
                "result": {
                  "text": "N/A"
                }
              },
              "type": "special"
            }
          ],
          "max": 100,
          "min": 0,
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "percent"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 7,
        "w": 5,
        "x": 1,
        "y": 0
      },
      "id": 12,
      "interval": null,
      "links": [],
      "maxDataPoints": 100,
      "options": {
        "orientation": "horizontal",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showThresholdLabels": false,
        "showThresholdMarkers": true,
        "text": {}
      },
      "pluginVersion": "8.2.7",
      "targets": [
        {
          "exemplar": true,
          "expr": "(pod_nvidia_gpu_memory_used_mib{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}/pod_nvidia_gpu_memory_total_mib{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"})*100",
          "format": "time_series",
          "interval": "",
          "intervalFactor": 1,
          "legendFormat": "",
          "refId": "A"
        }
      ],
      "title": "Memory Utilization",
      "transparent": true,
      "type": "gauge"
    },
    {
      "cacheTimeout": null,
      "datasource": "${datasource}",
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [
            {
              "options": {
                "match": "null",
                "result": {
                  "text": "N/A"
                }
              },
              "type": "special"
            }
          ],
          "max": 100,
          "min": 0,
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "red",
                "value": 80
              }
            ]
          },
          "unit": "percent"
        },
        "overrides": []
      },
      "gridPos": {
        "h": 7,
        "w": 5,
        "x": 6,
        "y": 0
      },
      "id": 14,
      "interval": null,
      "links": [],
      "maxDataPoints": 100,
      "options": {
        "orientation": "horizontal",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "showThresholdLabels": false,
        "showThresholdMarkers": true,
        "text": {}
      },
      "pluginVersion": "8.2.7",
      "targets": [
        {
          "exemplar": true,
          "expr": "pod_nvidia_gpu_duty_cycle{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}",
          "format": "time_series",
          "interval": "",
          "intervalFactor": 1,
          "legendFormat": "",
          "refId": "A"
        }
      ],
      "title": "GPU Utilization",
      "transparent": true,
      "type": "gauge"
    },
    {
      "aliasColors": {},
      "bars": false,
      "dashLength": 10,
      "dashes": false,
      "datasource": "prom-16fnzvzp",
      "fill": 0,
      "fillGradient": 0,
      "gridPos": {
        "h": 7,
        "w": 12,
        "x": 12,
        "y": 0
      },
      "hiddenSeries": false,
      "id": 10,
      "legend": {
        "alignAsTable": false,
        "avg": false,
        "current": false,
        "hideZero": false,
        "max": false,
        "min": false,
        "rightSide": false,
        "show": true,
        "total": false,
        "values": false
      },
      "lines": true,
      "linewidth": 1,
      "links": [],
      "nullPointMode": "null",
      "options": {
        "alertThreshold": true
      },
      "percentage": false,
      "pluginVersion": "8.2.7",
      "pointradius": 5,
      "points": false,
      "renderer": "flot",
      "seriesOverrides": [
        {
          "$$hashKey": "object:1468"
        }
      ],
      "spaceLength": 10,
      "stack": false,
      "steppedLine": false,
      "targets": [
        {
          "exemplar": true,
          "expr": "pod_nvidia_gpu_duty_cycle{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}",
          "format": "time_series",
          "interval": "",
          "intervalFactor": 1,
          "legendFormat": "",
          "refId": "A"
        }
      ],
      "thresholds": [],
      "timeFrom": null,
      "timeRegions": [],
      "timeShift": null,
      "title": "GPU Utilization / Card",
      "tooltip": {
        "shared": true,
        "sort": 2,
        "value_type": "individual"
      },
      "transparent": true,
      "type": "graph",
      "xaxis": {
        "buckets": null,
        "mode": "time",
        "name": null,
        "show": true,
        "values": []
      },
      "yaxes": [
        {
          "$$hashKey": "object:73",
          "format": "percent",
          "label": null,
          "logBase": 1,
          "max": null,
          "min": null,
          "show": true
        },
        {
          "$$hashKey": "object:74",
          "format": "short",
          "label": null,
          "logBase": 1,
          "max": null,
          "min": null,
          "show": true
        }
      ],
      "yaxis": {
        "align": false,
        "alignLevel": null
      }
    },
    {
      "aliasColors": {},
      "bars": false,
      "dashLength": 10,
      "dashes": false,
      "datasource": "${datasource}",
      "fill": 0,
      "fillGradient": 0,
      "gridPos": {
        "h": 9,
        "w": 11,
        "x": 0,
        "y": 7
      },
      "hiddenSeries": false,
      "id": 8,
      "legend": {
        "alignAsTable": false,
        "avg": false,
        "current": false,
        "hideZero": false,
        "max": false,
        "min": false,
        "rightSide": false,
        "show": true,
        "total": false,
        "values": false
      },
      "lines": true,
      "linewidth": 1,
      "links": [],
      "nullPointMode": "null",
      "options": {
        "alertThreshold": true
      },
      "percentage": false,
      "pluginVersion": "8.2.7",
      "pointradius": 5,
      "points": false,
      "renderer": "flot",
      "seriesOverrides": [
        {
          "$$hashKey": "object:859"
        }
      ],
      "spaceLength": 10,
      "stack": false,
      "steppedLine": false,
      "targets": [
        {
          "exemplar": true,
          "expr": "(pod_nvidia_gpu_memory_used_mib{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}/pod_nvidia_gpu_memory_total_mib{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"})*100",
          "format": "time_series",
          "interval": "",
          "intervalFactor": 1,
          "legendFormat": "",
          "refId": "A"
        }
      ],
      "thresholds": [],
      "timeFrom": null,
      "timeRegions": [],
      "timeShift": null,
      "title": "Memory Utilization / Card",
      "tooltip": {
        "shared": true,
        "sort": 2,
        "value_type": "individual"
      },
      "transparent": true,
      "type": "graph",
      "xaxis": {
        "buckets": null,
        "mode": "time",
        "name": null,
        "show": true,
        "values": []
      },
      "yaxes": [
        {
          "$$hashKey": "object:229",
          "format": "percent",
          "label": null,
          "logBase": 1,
          "max": null,
          "min": null,
          "show": true
        },
        {
          "$$hashKey": "object:230",
          "format": "short",
          "label": null,
          "logBase": 1,
          "max": null,
          "min": null,
          "show": true
        }
      ],
      "yaxis": {
        "align": false,
        "alignLevel": null
      }
    },
    {
      "aliasColors": {},
      "bars": false,
      "dashLength": 10,
      "dashes": false,
      "datasource": "${datasource}",
      "fill": 0,
      "fillGradient": 0,
      "gridPos": {
        "h": 9,
        "w": 12,
        "x": 12,
        "y": 7
      },
      "hiddenSeries": false,
      "id": 2,
      "legend": {
        "alignAsTable": false,
        "avg": false,
        "current": false,
        "hideZero": false,
        "max": false,
        "min": false,
        "rightSide": false,
        "show": true,
        "total": false,
        "values": false
      },
      "lines": true,
      "linewidth": 1,
      "links": [],
      "nullPointMode": "null",
      "options": {
        "alertThreshold": true
      },
      "percentage": false,
      "pluginVersion": "8.2.7",
      "pointradius": 5,
      "points": false,
      "renderer": "flot",
      "seriesOverrides": [],
      "spaceLength": 10,
      "stack": false,
      "steppedLine": false,
      "targets": [
        {
          "exemplar": true,
          "expr": "pod_nvidia_gpu_memory_used_mib{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}",
          "format": "time_series",
          "interval": "",
          "intervalFactor": 1,
          "legendFormat": "",
          "refId": "A"
        }
      ],
      "thresholds": [],
      "timeFrom": null,
      "timeRegions": [],
      "timeShift": null,
      "title": "Memory Usage/ Card",
      "tooltip": {
        "shared": true,
        "sort": 2,
        "value_type": "individual"
      },
      "transparent": true,
      "type": "graph",
      "xaxis": {
        "buckets": null,
        "mode": "time",
        "name": null,
        "show": true,
        "values": []
      },
      "yaxes": [
        {
          "$$hashKey": "object:308",
          "format": "mbytes",
          "label": null,
          "logBase": 1,
          "max": null,
          "min": null,
          "show": true
        },
        {
          "$$hashKey": "object:309",
          "format": "short",
          "label": null,
          "logBase": 1,
          "max": null,
          "min": null,
          "show": true
        }
      ],
      "yaxis": {
        "align": false,
        "alignLevel": null
      }
    },
    {
      "aliasColors": {},
      "bars": false,
      "dashLength": 10,
      "dashes": false,
      "datasource": "${datasource}",
      "fill": 4,
      "fillGradient": 0,
      "gridPos": {
        "h": 9,
        "w": 24,
        "x": 0,
        "y": 16
      },
      "hiddenSeries": false,
      "id": 6,
      "legend": {
        "alignAsTable": true,
        "avg": false,
        "current": false,
        "max": false,
        "min": false,
        "rightSide": false,
        "show": true,
        "total": false,
        "values": false
      },
      "lines": true,
      "linewidth": 1,
      "links": [],
      "nullPointMode": "null",
      "options": {
        "alertThreshold": true
      },
      "percentage": false,
      "pluginVersion": "8.2.7",
      "pointradius": 5,
      "points": false,
      "renderer": "flot",
      "seriesOverrides": [],
      "spaceLength": 10,
      "stack": true,
      "steppedLine": false,
      "targets": [
        {
          "exemplar": true,
          "expr": "pod_nvidia_gpu_power_usage_watts{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}",
          "format": "time_series",
          "interval": "",
          "intervalFactor": 1,
          "legendFormat": "",
          "refId": "A"
        }
      ],
      "thresholds": [],
      "timeFrom": null,
      "timeRegions": [],
      "timeShift": null,
      "title": "Power Usage",
      "tooltip": {
        "shared": true,
        "sort": 2,
        "value_type": "individual"
      },
      "type": "graph",
      "xaxis": {
        "buckets": null,
        "mode": "time",
        "name": null,
        "show": true,
        "values": []
      },
      "yaxes": [
        {
          "$$hashKey": "object:701",
          "format": "watt",
          "label": null,
          "logBase": 1,
          "max": null,
          "min": null,
          "show": true
        },
        {
          "$$hashKey": "object:702",
          "format": "short",
          "label": null,
          "logBase": 1,
          "max": null,
          "min": null,
          "show": true
        }
      ],
      "yaxis": {
        "align": false,
        "alignLevel": null
      }
    }
  ],
  "refresh": false,
  "schemaVersion": 32,
  "style": "dark",
  "tags": [],
  "templating": {
    "list": [
      {
        "description": null,
        "error": null,
        "hide": 0,
        "includeAll": false,
        "label": null,
        "multi": false,
        "name": "datasource",
        "options": [],
        "query": "prometheus",
        "refresh": 1,
        "regex": "",
        "skipUrlSync": false,
        "type": "datasource"
      },
      {
        "allValue": null,
        "datasource": "${datasource}",
        "definition": "label_values(pod_nvidia_gpu_num_devices, cluster)",
        "description": null,
        "error": null,
        "hide": 0,
        "includeAll": false,
        "label": null,
        "multi": false,
        "name": "cluster",
        "options": [],
        "query": {
          "query": "label_values(pod_nvidia_gpu_num_devices, cluster)",
          "refId": "StandardVariableQuery"
        },
        "refresh": 1,
        "regex": "",
        "skipUrlSync": false,
        "sort": 1,
        "type": "query"
      },
      {
        "allValue": null,
        "datasource": "${datasource}",
        "definition": "label_values(pod_nvidia_gpu_num_devices{cluster=\"$cluster\"}, namespace)",
        "description": null,
        "error": null,
        "hide": 0,
        "includeAll": false,
        "label": null,
        "multi": false,
        "name": "namespace",
        "options": [],
        "query": {
          "query": "label_values(pod_nvidia_gpu_num_devices{cluster=\"$cluster\"}, namespace)",
          "refId": "StandardVariableQuery"
        },
        "refresh": 1,
        "regex": "",
        "skipUrlSync": false,
        "sort": 0,
        "type": "query"
      },
      {
        "allValue": null,
        "datasource": "${datasource}",
        "definition": "label_values(pod_nvidia_gpu_num_devices{cluster=\"$cluster\", namespace=\"$namespace\"}, pod)",
        "description": null,
        "error": null,
        "hide": 0,
        "includeAll": false,
        "label": null,
        "multi": false,
        "name": "pod",
        "options": [],
        "query": {
          "query": "label_values(pod_nvidia_gpu_num_devices{cluster=\"$cluster\", namespace=\"$namespace\"}, pod)",
          "refId": "StandardVariableQuery"
        },
        "refresh": 1,
        "regex": "",
        "skipUrlSync": false,
        "sort": 1,
        "type": "query"
      }
    ]
  },
  "time": {
    "from": "now-1h",
    "to": "now"
  },
  "timepicker": {},
  "timezone": "",
  "title": "GPU / pod-eks",
  "uid": "37tllT74z",
  "version": 5
}

Grafana 面板效果展示（部分）

Grafana 效果

【数据结构】算法效率的度量方法
上一篇
2023人工智能工程五大新方向
下一篇