如何在CloudWatch中按单台服务器展示统计数据而非整体平均值？

我已经把JVM线程数、CPU使用率等指标发布到了CloudWatch Logs（后续转为CloudWatch Metrics可能更合适）。

基础查询语句 #

fields @timestamp, jvm.threads.live, process.cpu.usage, server , system.load.average.1m , jvm.memory.usage.after.gc
| sort by @timestamp desc

示例数据 #

{
    "@timestamp": "2022-xx-xxT11:03:00.027Z",
    "jvm.gc.overhead": "0",
    "jvm.memory.usage.after.gc": "0.036",
    "jvm.threads.live": "47",
    "process.cpu.usage": "1",
    "server": "MyServer.localdomain",
    "system.load.average.1m": "2.351"
}

之前用以下语句生成图表，但它会计算所有服务器的平均值，没有区分单台服务器：

fields @timestamp, jvm.threads.live, process.cpu.usage, server , system.load.average.1m , jvm.memory.usage.after.gc
| stats avg(jvm.threads.live) as threads, avg(process.cpu.usage * 100) as cpuPercentage, avg(system.load.average.1m) as loadAvg, avg(jvm.memory.usage.after.gc * 100) as memoryPercentage  by bin(1m)

因为服务器名称是随机生成的，无法提前获取完整列表，要展示每台服务器的单独统计数据，只需在stats语句的by子句中加入server字段即可：

fields @timestamp, jvm.threads.live, process.cpu.usage, server , system.load.average.1m , jvm.memory.usage.after.gc
| stats avg(jvm.threads.live) as threads, avg(process.cpu.usage * 100) as cpuPercentage, avg(system.load.average.1m) as loadAvg, avg(jvm.memory.usage.after.gc * 100) as memoryPercentage  by bin(1m), server

修改后的查询会按1分钟时间分片和服务器名称双重分组，自动为每台服务器计算对应时间区间内的指标平均值，生成图表时也会按server字段区分不同系列，无需提前知晓服务器列表。

内容的提问来源于stack exchange，提问作者rjdkolb