Cloudwatch-to-Graphite

Cloudwatch-to-Graphite (leadbutt) is a small utility to take metrics from CloudWatch to Graphite.

Installation

Install using pip:

pip install cloudwatch-to-graphite

Configuring boto

Cloudwatch-to-Graphite uses boto, so make sure to follow its configuration instructions. The easiest way to do this is to set up the AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY environment variables.

Usage

Configuration Files

If you have a simple setup, the easiest way to get started is to set up a config.yaml. You can copy the included config.yaml.example. Then just run:

leadbutt

If you have several configs you want to switch between, you can specify a custom configuration file:

leadbutt --config-file=production.yaml -n 20

You can even generate configs on the fly and send them in via stdin by setting the config file to '-':

generate_config_from_inventory | leadbutt --config-file=-

There's a helper to generate configuration files called plumbum. Use it like:

plumbum [-r REGION] [-f FILTER] [--token TOKEN] template namespace

Namespace is the CloudWatch namespace for the resources of interest; for example AWS/RDS. The template is a Jinja2 template. You can add arbitrary replacement tokens, eg {{ replace_me }}, and then pass in values on the CLI via --token. For example, if you called:

plumbum --token replace_me='hello, world' sample_templates/rds.yml.j2 AWS/RDS

You would get all instances of {{ replace_me }} in the templace replaced with hello, world.

Filters

You can pass simple key=value filters in to plumbum; be aware of the limitations:

• the filters run against whatever the AWS API has returned; if you have a lot of objects of whatever type, expect the API request to take a while.
• they work only against object attributes and tags returned by the API. For example, RDS and ELB objects can be tagged, but as getting the tags is a per-object subrequest; plumbum does not do those, so you can only filter on the object attributes.

Example: plumbum -f Name=my-dev-instance sample_templates/ec2.yml.j2 ec2

Sending Data to Graphite

If your graphite server is at graphite.local, you can send metrics by chaining with netcat:

leadbutt | nc -q0 graphite.local 2003

Or if you want to use UDP:

leadbutt | nc -uw0 graphite.local 2003

If you need to namespace your metrics for a hosted Graphite provider, you could provide a custom formatter, but the easiest way is to just run the output through awk:

leadbutt | \
  awk -v namespace="$HOSTEDGRAPHITE_APIKEY" '{print namespace"."$0}' | \
  nc -uw0 my-graphite-provider.xxx 2003

Customizing Your Graphite Metric Names

Set the Formatter option to set the template used to generate Graphite metric names. I wasn't sure what should be default, so I copied cloudwatch2graphite's. Here's what it looks like:

cloudwatch.%(Namespace)s.%(dimension)s.%(MetricName)s.%(statistic)s.%(Unit)s

TitleCased variables come directly from the YAML configuration, while lowercase variables are derived:

• statistic -- the current statistic since Statistics can be a list
• dimension -- the dimension value, e.g. "i-r0b0t" or "my-load-balancer"

The format string is Python's %-style.

config.yaml

What metrics are pulled is in a YAML configuration file. See the example config.yaml.example for an idea of what you can do.

Developing

See: : Contributing.

Useful References

• CloudWatch Reference
• boto CloudWatch docs

Prior Art

Cloudwatch-to-Graphite was inspired by edasque's cloudwatch2graphite. I was looking to expand it, but I wanted to use boto.