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
Data-Miner16 天前
大语言模型+智能体AI,122页PPT详解落地应用培训!
人工智能·microsoft·语言模型
c++之路16 天前
备忘录模式(Memento Pattern)
c++·microsoft
Solis程序员16 天前
MCP (Model Context Protocol):AI应用连接外部世界的标准协议
人工智能·microsoft·agent·skill·mcp
诺未科技_NovaTech16 天前
上海诺未携手惠灵顿中国,基于微软 Azure 打造 AI 教育生态标杆
人工智能·microsoft·azure·ai教育
hnult16 天前
在线笔试平台如何选型?考试云九重防作弊 + 六大 AI 能力 智能招聘笔试解决方案
人工智能·笔记·microsoft·信息可视化·课程设计
步步为营DotNet16 天前
Blazor 与 Microsoft.Extensions.AI 在客户端性能优化中的协同应用
人工智能·microsoft·性能优化
hai31524754316 天前
九章编程法 · 猜数字游戏 (GW-BASIC 重构版) *
人工智能·microsoft·游戏引擎·游戏程序
weixin_3975740917 天前
从AI问答到AI执行:企业智能体平台的定位跃迁
人工智能·microsoft
basketball61617 天前
AI Infra 硬件体系与编程模型:17. CUDA编程基础:底层驱动 API 调用
人工智能·microsoft·nvidia·cuda