Azure AKS集群监控告警表达式配置

背景需求

Azure AKS集群中,需要对部署的服务进行监控和告警,需要创建并启用预警规则,而这里怎么去监控每个pod级别的CPU和内存,需要自己写搜索查询

解决方法

搜索和查询的语句如下,需要自己替换其中的部分信息,其中的clusterID需要在AKS集群中的概述-右侧JSON视图-id,其中id就是我们要的clusterID,然后替换下面的clusterID字段内容。最后的 percentage 也需要根据实际情况来设置对应值。

复制代码
#内存
let metricUsageCounterName = 'memoryRssBytes';
let metricLimitCounterName = 'memoryLimitBytes';
let clusterID = "/subscriptions/xxxxxxxxxxx";
let CachedFilteredPerfTable = Perf
| where ObjectName == 'K8SContainer'
| where InstanceName startswith clusterID
| project Node = Computer, TimeGenerated, CounterName, CounterValue, InstanceName;
let LimitsTable = CachedFilteredPerfTable
| where CounterName =~ metricLimitCounterName
| summarize arg_max(TimeGenerated, *) by Node, InstanceName
| project Node, InstanceName, LimitsValue = CounterValue, TimeGenerated;
let MetaDataTable = KubePodInventory
| where isnotempty(ClusterName) | where isnotempty(Namespace) | where isnotempty(Computer)
| where ClusterId =~ clusterID
| project TimeGenerated, ClusterId, Namespace, ControllerName, Node = Computer, Pod = Name, ContainerInstance = ContainerName, ContainerID, ControllerKind = ControllerKind
| summarize arg_max(TimeGenerated, *) by Node, ContainerInstance
| project Namespace, ControllerName, Node, Pod, ContainerInstance, InstanceName = strcat(ClusterId, '/', ContainerInstance), ContainerID, ControllerKind, 
ContainerName = tostring(split(ContainerInstance, '/')[1]), LastPodInventoryTimeGenerated = TimeGenerated, ClusterId
| join kind=leftouter (LimitsTable) on Node, InstanceName
| project Namespace, ControllerName, Node, Pod, InstanceName, ContainerID, LimitsValue, ControllerKind, ContainerName, ContainerInstance, LastPodInventoryTimeGenerated, ClusterId;
let AggregationTable = CachedFilteredPerfTable
| where CounterName =~ metricUsageCounterName
| project TimeGenerated, Node, InstanceName, CounterValue
| summarize  Aggregation = percentile(CounterValue, 95) by Node, InstanceName 
| project Node, InstanceName, Aggregation;
MetaDataTable
| join kind= leftouter( AggregationTable ) on Node, InstanceName
| order by ContainerName asc, ContainerName
| extend ContainerIdentity = strcat(ContainerName, '|', Pod)
| extend percentage = Aggregation/LimitsValue * 100
| project ContainerIdentity, percentage, Aggregation, LimitsValue, Node, ControllerName, ControllerKind, ContainerID, ContainerInstance, InstanceName, Namespace, LastPodInventoryTimeGenerated, ClusterId
| where percentage > 80






#CPU
let metricUsageCounterName = 'cpuUsageNanoCores';
let metricLimitCounterName = 'cpuLimitNanoCores'; 
let clusterID = "/subscriptions/xxxxxxxxxxx";
let CachedFilteredPerfTable = Perf
| where ObjectName == 'K8SContainer'
| where InstanceName startswith clusterID
| project Node = Computer, TimeGenerated, CounterName, CounterValue, InstanceName;
let LimitsTable = CachedFilteredPerfTable
| where CounterName =~ metricLimitCounterName
| summarize arg_max(TimeGenerated, *) by Node, InstanceName
| project Node, InstanceName, LimitsValue = CounterValue/1000000, TimeGenerated;
let MetaDataTable = KubePodInventory
| where isnotempty(ClusterName) | where isnotempty(Namespace) | where isnotempty(Computer)
| where ClusterId =~ clusterID
| project TimeGenerated, ClusterId, Namespace, ControllerName, Node = Computer, Pod = Name, ContainerInstance = ContainerName, ContainerID, ControllerKind = ControllerKind
| summarize arg_max(TimeGenerated, *) by Node, ContainerInstance
| project Namespace, ControllerName, Node, Pod, ContainerInstance, InstanceName = strcat(ClusterId, '/', ContainerInstance), ContainerID, ControllerKind, 
ContainerName = tostring(split(ContainerInstance, '/')[1]), LastPodInventoryTimeGenerated = TimeGenerated, ClusterId
| join kind=leftouter (LimitsTable) on Node, InstanceName
| project Namespace, ControllerName, Node, Pod, InstanceName, ContainerID, LimitsValue, ControllerKind, ContainerName, ContainerInstance, LastPodInventoryTimeGenerated, ClusterId;
let AggregationTable = CachedFilteredPerfTable
| where CounterName =~ metricUsageCounterName
| project TimeGenerated, Node, InstanceName, CounterValue = CounterValue/1000000
| summarize  Aggregation = percentile(CounterValue, 95) by Node, InstanceName 
| project Node, InstanceName, Aggregation;
MetaDataTable
| join kind= leftouter( AggregationTable ) on Node, InstanceName
| order by ContainerName asc, ContainerName
| extend ContainerIdentity = strcat(ContainerName, '|', Pod)
| extend percentage = Aggregation/LimitsValue * 100
| project ContainerIdentity, percentage, Aggregation, LimitsValue, Node, ControllerName, ControllerKind, ContainerID, ContainerInstance, InstanceName, Namespace, LastPodInventoryTimeGenerated, ClusterId
| where percentage > 80
相关推荐
会飞的小新20 小时前
C 标准库之 <fenv.h> 详解与深度解析
c语言·开发语言·microsoft
编码者卢布21 小时前
【Azure Policy】Policy修正任务为Azure资源添加诊断日志报错问题的调查
microsoft·flask·azure
海盗123421 小时前
微软技术周报——2026-07-22
microsoft·c#·.net
海盗123421 小时前
微软技术日报 ——2026-07-21
microsoft·c#·.net
半亩码田1 天前
【.NET新特性·第8篇】.NET 9 AI 构建基块:Microsoft.Extensions.AI
人工智能·microsoft·.net
编码者卢布3 天前
【Azure APIM】APIM的诊断日志与Application Insights的日志是否可以串联为一个端到端的日志链路呢?
python·flask·azure
老大白菜4 天前
Omnigent(omni)使用说明
服务器·数据库·microsoft
hkj88085 天前
Linux 总线-设备-驱动(Bus-Device-Driver)完整协作原理
linux·运维·microsoft
facaixxx20245 天前
AI模型的Function Calling能力:工作原理及模型支持清单
microsoft
观远数据6 天前
从离线开发到实时同步:DataFlow如何支撑企业级数据治理闭环
java·windows·microsoft·excel