根据Kusto中​​的键合并来自多个表的数据

时间:2020-09-23 16:32:49

标签: azure-log-analytics kusto kusto-query-language

我正在尝试在Azure Log Analytics中合并多个表。每个表都有一个唯一列和一个公共列。将它们与Join()合并效率不高,因为我一次只能创建两个表。 Union()似乎是正确的函数,但是当我合并表时,在公共列中以重复的行结束。

示例:

// CPU usage
let CPU_table=VPN_Metrics_CL | extend timestamp = (todatetime(ts_s)+7h) 
| where metric_s == "system/cpmCPUTotal1Min.rrd"  
| extend region = substring(host_s,0,4)
| summarize maxCPU = max(val_d) by region
| extend score_CPU = case(maxCPU <= 59, 0, 
                          maxCPU <= 79, 1, 
                          3)
| project score_CPU, region;
// Memory usage
let Memory_table=VPN_Metrics_CL| extend timestamp = todatetime(ts_s)+7h
| where metric_s in ("hw_mem_used_pct") and val_d >= 0 and host_s contains "vpn"
| extend region = substring(host_s,0,4)
| summarize maxMemory = max(val_d) by region
| extend score_mem = case(maxMemory <= 59, 0, 
                          maxMemory <= 79, 1, 
                          3)
| project score_mem, region;
union CPU_table, Memory_table

我计划总共拥有10张以上的桌子。

这是结果:

score_mem  |  score_CPU  |  region
    0                         USA
                 0            USA
  etc. etc.

如何基于键合并行?关键是区域。

谢谢

2 个答案:

答案 0 :(得分:1)

如果源是同一张表,则最有效的方法是使用条件聚合:

let isCpuMetric = (metric_s:string) {metric_s == "system/cpmCPUTotal1Min.rrd"};
let isMemoryMetric =  (metric_s:string, val_d:double, host_s:string) {metric_s in ("hw_mem_used_pct") and val_d >= 0 and host_s contains "vpn"};
VPN_Metrics_CL 
| extend timestamp = (todatetime(ts_s)+7h) 
| extend region = substring(host_s,0,4)
| where isCpuMetric(metric_s) or isMemoryMetric(metric_s, val_d, host_s)
| summarize maxCPU = maxif(val_d, isCpuMetric(metric_s)), maxMemory=maxif(val_d, isMemoryMetric(metric_s, val_d, host_s)) by region
| extend score_mem = case(maxMemory <= 59, 0, maxMemory <= 79, 1, 3),
         score_CPU = case(maxCPU <= 59, 0, maxCPU <= 79, 1, 3)

如果来源不同-您仍然可以加入或查找运算符。如果您有结果R1 .. RN-来自子查询:

R1 
| lookup R2 on Region
| lookup R3 on Region
...

查找运算符的文档:https://docs.microsoft.com/en-us/azure/data-explorer/kusto/query/lookupoperator

答案 1 :(得分:0)

我发现为每个类别的分数列赋予相同的名称更加容易:“得分” 然后使用Union,我合并所有表格并汇总总分。

union CPU_table, Memory_table, AAA_table, bw_data, more_tables.....
| summarize score_total = sum(score) by region, bin(timestamp, $__interval)
| project score_total, region, timestamp