Configuring Recording Rules to Improve Metric Query Efficiency

Updated on 2025-03-03 GMT+08:00

View PDF

Recording rules can be used for secondary development of metric data. By setting recording rules, you can move the computing process to the write end, reducing resource usage on the query end.

Scenario

Some metrics may require much calculation on the query end, affecting query performance. You can configure recording rules to calculate common or complex metrics in advance. Especially in large-scale clusters and complex service scenarios, recording rules can reduce PromQL complexity, improve metric query performance, and prevent slow configuration and query.

Prerequisite

Both your service and CCE cluster have been connected to a Prometheus instance for CCE. For details, see Using Prometheus Monitoring to Monitor CCE Cluster Metrics.

Configuring a Recording Rule

Log in to the AOM 2.0 console.
In the navigation pane on the left, choose Prometheus Monitoring > Instances.
In the instance list, click a Prometheus instance for CCE.
In the navigation pane on the left, choose Settings. In the Recording Rules area, click Edit RecordingRule.yaml.

In the dialog box that is displayed, delete the default content and enter a custom recording rule.

Only one RecordingRule.yaml file needs to be configured for a cluster. Each rule group name must be unique.

Figure 1 Configuring a recording rule

**Table 1** Recording rule parameters
Parameter	Description
groups	Rule group. You can set multiple rule groups in one RecordingRule.yaml file.
name	Rule group name. Each rule group name must be unique.
interval	Execution interval of a rule group. The default value is 60s. (Optional)
rules	Rule. A rule group can contain multiple rules.
record	Name of a rule. The name must comply with Prometheus metric name specifications.
expr	Calculation expression. It is used to calculate metric values. It must comply with PromQL requirements.
labels	Metric label. Labels must comply with Prometheus metric label specifications. (Optional)

Example of a recording rule:

groups:
  - name: apiserver_request_total
    interval: 60s
    rules:
      - record: apiserver_request_rate
        expr: avg by (job, instance, mode) (rate(apiserver_request_total[5m]))
        labels:
          team: operations
      - record: job:apiserver_request_total:sum_rate10m
        expr: sum by (job)(rate(apiserver_request_total[10m]))
        labels:
          team: operations