isabella232 / dropwizard-librato

Send metrics from your Dropwizard application to Librato Metrics

DEPRECATED: If you are using AppOptics please refer to dropwizard-metrics-appoptics.

Overview

This project provides an easy way to send metrics from your Dropwizard project to Librato. It automatically includes the metrics-librato library and sets up the reporter based on your application YAML config. For more information on how you can use metrics-librato at runtime, please see some example usages.

Usage

There are two steps. First, you must add the dropwizard-metrics-librato Maven dependency to your POM file. Second, the application config YAML needs to be updated to configure the Librato Reporter, which will send your Metrics data to Librato.

First, add the dropwizard-metrics-librato dependency in your POM:

For Dropwizard 1.x and greater

<dependency>
    <groupId>com.librato.metrics</groupId>
    <artifactId>dropwizard-metrics-librato</artifactId>
    <version>10.2.0.4</version>
</dependency>

For earlier than Dropwizard 1.x

<dependency>
    <groupId>com.librato.metrics</groupId>
    <artifactId>dropwizard-metrics-librato</artifactId>
    <version>1.9.1.12</version>
</dependency>

Next, add a metrics configuration element to your YAML config file:

metrics:
  reporters:
    - type: librato
      username: "<Librato Email>"
      token: "<Librato API Token>"
      source: "<Source Identifier (usually hostname)>"
      timeout: [optional (int), number of seconds, defaults to 5]
      prefix: [optional (string), prepended to metric names]
      name: [optional (string), name of the reporter]

That's it. Once your application starts, your metrics should soon appear in Librato.

Environment Variables

If you wish to not have your username, token, or source in the configuration, you can alternatively supply them by setting the LIBRATO_USERNAME, LIBRATO_TOKEN, and LIBRATO_SOURCE environment variables, respectively

Tagging Support

If you'd like to participate in our tagging beta, please email [email protected] to be added. Once this is done, you may update your yaml file:

metrics:
  reporters:
    - type: librato
      username: "<Librato Email>"
      token: "<Librato API Token>"
      source: "<Source Identifier (usually hostname)>"
      timeout: [optional (int), number of seconds, defaults to 5]
      prefix: [optional (string), prepended to metric names]
      name: [optional (string), name of the reporter]
      tags:
        enabled: true
        static:
            service: "my-auth-service"
            environment: "staging"
        environment:
            node: NODE_NAME

The static tags are completely defined in the yaml file. The environment tags' names are also defined in the yaml file, but their values are determined by the environment variables at the time the Librato reporter is created. In this case, the environment variable with the name NODE_NAME would be queried and then assigned to the tag name node.

Whitelist / Blacklist

By default, all expanded metrics (percentiles, rates) are submitted for each Timer, Histogram, and Meter. If you wish to whitelist only certain metrics, you can do so like this:

metrics:
  reporters:
    - type: librato
      username: "<Librato Email>"
      token: "<Librato API Token>"
      source: "<Source Identifier (usually hostname)>"
      timeout: [optional (int), number of seconds, defaults to 5]
      prefix: [optional (string), prepended to metric names]
      name: [optional (string), name of the reporter]
      metricWhitelist:
      	- PCT_75
      	- PCT_98
      	- PCT_99
      	- RATE_MEAN
      	- RATE_1_MINUTE
      	- RATE_5_MINUTE

Similarly, if you wish to blacklist certain expanded metrics, you would do something similar to the above example, but replace metricWhitelist with metricBlacklist.

The full set of expanded metric names that you can specify are:

• MEDIAN
• PCT_75
• PCT_95
• PCT_98
• PCT_99
• PCT_999
• COUNT
• RATE_MEAN
• RATE_1_MINUTE
• RATE_5_MINUTE
• RATE_15_MINUTE

Note that you cannot supply both metricWhitelist and metricBlacklist.

Notes

For each logical application, it is highly recommend you use a prefix to distinguish common sets of metrics.

Contributors

• Initial code: Chris Huang