Prometheus Cheatsheet

PromQL (Prometheus Query Language) uses a simple expression language to select and aggregate time series data. Understanding the basic syntax is essential for building effective queries.

Prometheus collects time-series data as metrics. Each metric type has specific characteristics that determine how data is collected, stored, and queried. Understanding these types is essential for writing effective PromQL queries.

Simple selectors are the core building blocks of PromQL queries, allowing you to select time series metrics based on their names and labels. Labels are key-value pairs attached to metrics that provide additional context about what the metric represents.

Range vectors select a range of samples for each time series over a specified time interval. They are essential for calculating rates, averages, and other time-based aggregations in Prometheus.

Aggregation operators in PromQL allow you to combine multiple time series, perform calculations across dimensions, and reduce the dimensionality of your data. They are essential for creating meaningful summaries from large sets of metrics.

Binary operators in PromQL work with two operands and enable arithmetic, comparison, and logical operations between instant vectors. Understanding these operators is essential for creating complex queries and alerts.

Rate functions are essential for analyzing counter metrics over time, allowing you to calculate the rate of change, detect spikes, and understand trends in your time series data.

Time functions in PromQL allow you to work with time-related operations and create time-based calculations. These functions are useful for creating dynamic time references, filtering data based on time periods, and creating time-based alerts.

PromQL provides functions to manipulate metric labels dynamically. These functions allow you to add, modify, or transform labels without changing the underlying data, which is useful for relabeling metrics, creating consistent naming schemes, or preparing data for visualization.

Histograms in Prometheus track the distribution of values by counting observations in configurable buckets. The histogram_quantile() function allows you to calculate quantiles (like p50, p90, p99) from these histogram metrics to analyze latency distributions and other performance characteristics.

The topk() and bottomk() functions allow you to identify the highest or lowest values in a set of time series. These functions are particularly useful for finding outliers, top consumers of resources, or least utilized components in your system.

Vector matching in PromQL allows you to combine different time series data using binary operators. There are several types of matching behaviors that control how time series with different label sets are combined.

Subqueries allow you to run an inner query over a range of time and then use the result as input to an outer query function. They're particularly useful for calculating rates of rates or applying aggregations over already aggregated data.

Common PromQL query patterns that are frequently used for monitoring, alerting, and dashboards. These patterns solve typical observability challenges and can be adapted to your specific metrics.

Recording rules allow you to precompute frequently used or computationally expensive PromQL expressions and save their result as a new time series. This improves query performance and reduces load on the Prometheus server.