What are Prometheus labels?

Prometheus labels are key-value pairs used to add metadata to metrics, making them more descriptive and allowing flexible queries. They enable Prometheus to organize, filter, and aggregate metrics efficiently.

Key characteristics of labels

1. Key-value format: Each label is a key paired with a value. Example: status="200".

2. Attached to metrics: Labels provide context or dimensions to raw metrics. For instance, the metric http_requests_total can have labels like method="GET" or endpoint="/api".

3. Dynamic values: Labels can represent dynamic attributes, such as instance names, environments, or HTTP status codes.

Example of labels

Metric with labels:

http_requests_total{method="GET", endpoint="/home", status="200"} 42

Breakdown:

• http_requests_total: The metric name.
• {method="GET", endpoint="/home", status="200"}: Labels that provide context.
• 42: The metric value.

Use cases for labels

1. Filtering metrics: Labels allow querying specific subsets of metrics.

   • Query: http_requests_total{status="200"} retrieves metrics for successful requests.

2. Aggregating metrics: Labels enable grouping or aggregating data.

   • Query: sum(rate(http_requests_total[5m])) by (status) calculates the rate of requests grouped by status.

3. Adding context: Labels provide metadata like instance, region, or version to distinguish metrics.

Best practices for using labels

1. Avoid high cardinality: Labels with many unique values (like user_id) can lead to a large number of time series, increasing resource usage.

2. Use meaningful keys: Use descriptive and consistent keys such as region, environment, or status.

3. Keep values short: Avoid embedding long strings or logs as label values.

Label pitfalls

1. Unbounded values: Dynamic values like UUIDs or timestamps can create infinite time series, overwhelming Prometheus.

2. Query complexity: Excessive labels can make PromQL queries difficult to construct and understand.

3. Immutable labels: Labels cannot be modified once a time series is created. Changes require creating a new time series.

Labels are essential for making Prometheus metrics flexible and insightful. When used effectively, they enable powerful filtering, aggregation, and monitoring capabilities. However, careful design is required to avoid performance issues.

Got an article suggestion? Let us know

Explore more

What Is A Bucket In Prometheus?

Questions

Limitations of Prometheus Labels

Questions

What is the Difference Between a Gauge and a Counter?

Questions

What Are The 4 Types Of Metrics In Prometheus

Questions

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.