# How Can I Group Labels in a Prometheus Query?

Grouping labels in a Prometheus query allows you to aggregate metrics based on specific labels, providing a way to analyze data across different dimensions. You can use the `by` clause in Prometheus Query Language (PromQL) to group your results according to one or more labels. Here’s how to do it effectively.

[ad-logs-small]

### Basic Syntax for Grouping

The general syntax for grouping in a Prometheus query looks like this:

```
aggregation_function(metric_name) by (label1, label2, ...)
```

### Step-by-Step Examples

### 1. Counting Total Requests by Method

Suppose you have a metric called `http_requests_total` that counts HTTP requests. If you want to count the total number of requests grouped by the `method` label (e.g., `GET`, `POST`), you can use:

```
sum(http_requests_total) by (method)
```

This query sums the total requests and groups the result by the `method` label, giving you the total number of requests for each HTTP method.

### 2. Average Response Time by Status

If you want to calculate the average response time grouped by the `status` label, and you have a metric called `http_response_time_seconds`, you can use:

```
avg(http_response_time_seconds) by (status)
```

This returns the average response time for each unique status code.

### 3. Grouping by Multiple Labels

You can also group by multiple labels. For instance, if you want to count the total requests grouped by both `method` and `status`, you can do:

```
sum(http_requests_total) by (method, status)
```

This query returns the total number of requests for each combination of HTTP method and status.

### Step 4: Using Without and Without Keywords

### 1. Using `without`

If you want to aggregate metrics while ignoring certain labels, you can use the `without` keyword. For example, if you want to sum total requests but ignore the `instance` label, you can write:

```
sum(http_requests_total) without (instance)
```

This aggregates the total requests across all instances.

### 2. Using `on` and `ignoring`

You can also combine groupings using `on` and `ignoring`. For example, if you want to group by `method` but ignore the `instance` label:

```
sum(http_requests_total) by (method)
  * on(method)
  sum(http_response_time_seconds) by (method)
```

This allows you to perform operations while maintaining focus on the `method` label.

[ad-uptime]