如何在 Promscale 的聚合查询中对标签进行分组

Posted 2023-02-16

【中文标题】如何在 Promscale 的聚合查询中对标签进行分组

【英文标题】：How to group on labels in aggregate query to Promscale

【发布时间】：2021-07-24 03:03:57

【问题描述】：

我正在创建从 Grafana 到 Promscale 的 SQL 查询。有指标和标签。我无法获得按某些标签进行分组的正确方法。我试过了：

SELECT time_bucket('$__interval', "time") AS "time",
       AVG("value") AS "used"
  FROM "disk_used_percent"
 WHERE $__timeFilter("time") AND
       "labels" ? ('host' == '$host_pg')
 GROUP BY 1, "labels" --> 'path'
 ORDER BY 1;

还有：

SELECT time_bucket('$__interval', "time") AS "time",
       AVG("value") AS "used"
  FROM "disk_used_percent"
 WHERE $__timeFilter("time") AND
       "labels" ? ('host' == '$host_pg')
 GROUP BY 1, "path_id"
 ORDER BY 1;

但分组似乎没有按预期工作。怎么了？相应的 PromQL 查询将是：

avg(disk_used_percenthost=~"$host_prom") by(path))

【参考方案1】：

您可以使用VAL("<label>_id") 进行分组：

SELECT time_bucket('$__interval', "time") AS "time",
       VAL("path_id") AS "path",
       AVG("value") AS "used"
  FROM "disk_used_percent"
 WHERE $__timeFilter("time") AND
       "labels" ? ('host' == '$host_pg')
 GROUP BY 1, 2
 ORDER BY 1;

旁注：还要避免在 Grafana 中使用 $__timeFilter("time") 模板宏，因为它会生成以下谓词：

"time" BETWEEN 'time range begin' AND 'time range end'

这可能是 problematic 在某些情况下。