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
可涵不会debug2 天前
LangChain 进阶之路:LangGraph 图工作流、有状态智能体完整入门教程
microsoft
Access开发易登软件5 天前
Access 复杂业务计算,SQL 写在 VBA 还是存储过程?
microsoft·vba·access·access开发
宝桥南山8 天前
Microsoft Fabric - 简单尝试一下Microsoft Fabric .NET SDK
microsoft·微软·.net·.netcore·powerbi·fabric
hqyjzsb8 天前
规划工商管理大学成长:搭建四层能力体系,重视高阶的 AI 能力建设
开发语言·人工智能·python·microsoft·职场和发展·数据挖掘·业界资讯
ai_finder8 天前
买卖点预警系统是怎么工作的?从自然语言到盯盘任务的一次工程拆解
人工智能·科技·microsoft·金融
m0_466525299 天前
从火柴人看护画面看云从科技生态企业的隐私保护实践
大数据·人工智能·科技·microsoft
大模型码小白9 天前
告别造假数据,直接连数据库查真实时序数据喂给 TimechoAI 大模型
java·数据库·人工智能·microsoft·架构
海盗123410 天前
微软技术日报 2026-09-14:Rust 升为微软一级语言,云业务重组为智能体与基础设施
开发语言·microsoft·rust
ManageEngineITSM10 天前
什么是IT服务连续性管理?灾难恢复与业务连续性一文讲清
数据库·microsoft·工单系统·变更管理
秦哈哈10 天前
【Hello Agents】学习笔记(二)
笔记·学习·microsoft