# TelemetryMetricsCloudwatch
[](https://github.com/bmuller/telemetry_metrics_cloudwatch/actions/workflows/ci.yml)
[](https://hex.pm/packages/telemetry_metrics_cloudwatch)
[](https://hexdocs.pm/telemetry_metrics_cloudwatch/)
This is a [Amazon CloudWatch](https://aws.amazon.com/cloudwatch/) Reporter for [`Telemetry.Metrics`](https://github.com/beam-telemetry/telemetry_metrics) definitions.
## Installation
To install `telemetry_metrics_cloudwatch`, just add an entry to your `mix.exs`:
```elixir
def deps do
[
{:telemetry_metrics_cloudwatch, "~> 0.3"}
]
end
```
(Check [Hex](https://hex.pm/packages/telemetry_metrics_cloudwatch) to make sure you're using an up-to-date version number.)
## Usage
Provide a list of metric definitions to the `init/2` function. It's recommended to
run TelemetryMetricsCloudwatch under a supervision tree, usually under Application.
```elixir
def start(_type, _args) do
# List all child processes to be supervised
children = [
{TelemetryMetricsCloudwatch, [metrics: metrics()]}
...
]
opts = [strategy: :one_for_one, name: ExampleApp.Supervisor]
Supervisor.start_link(children, opts)
end
defp metrics do
[
counter("http.request.count"),
last_value("vm.memory.total", unit: :byte),
last_value("vm.total_run_queue_lengths.total")
]
end
```
You can also provide options for the namespace used in CloudWatch (by default, "Telemetry")
and the minimum frequency (in milliseconds) with which data will be posted (see section
below for posting rules). For instance:
```elixir
...
children = [
{TelemetryMetricsCloudwatch, metrics: metrics(), namespace: "Backend", push_interval: 30_000}
]
...
```
### Telemetry.Metrics Types Supported
`TelemetryMetricsCloudwatch` supports 4 of the [Metrics](https://hexdocs.pm/telemetry_metrics/Telemetry.Metrics.html#module-metrics):
* [Counter](https://hexdocs.pm/telemetry_metrics/Telemetry.Metrics.html#counter/2):
Counter metric keeps track of the total number of specific events emitted.
* [LastValue](https://hexdocs.pm/telemetry_metrics/Telemetry.Metrics.html#last_value/2):
Last value keeps track of the selected measurement found in the most recent event.
* [Summary](https://hexdocs.pm/telemetry_metrics/Telemetry.Metrics.html#summary/2): Summary
aggregates measurement's values into statistics, e.g. minimum and maximum, mean, or percentiles.
This sends every measurement to CloudWatch.
* [Sum](https://hexdocs.pm/telemetry_metrics/Telemetry.Metrics.html#sum/2): Sum metric keeps track
of the sum of selected measurement's values carried by specific events. If you are using Summary
for a metric already, then CloudWatch can calculate a Sum using that Summary metric. If you
only need a Sum (and no other summary metrics) then use this Sum metric instead.
These metrics are sent to CloudWatch based on the rules described below.
To write [high-resolution metrics](https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/publishingMetrics.html#high-resolution-metrics=), supply the `:storage_resolution` option (which can be the default of `:standard` or `:high`):
```elixir
counter(
"http.request.count",
reporter_options: [storage_resolution: :high]
)
```
### When Data is Sent
Cloudwatch has [certain constraints](https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/publishingMetrics.html)
on the number of metrics that can be sent up at any given time. `TelemetryMetricsCloudwatch`
will send accumulated metric data at least every minute (configurable by the `:push_interval`
option) or when the data cache has reached the maximum size that CloudWatch will accept.
### Event Sampling
You can optionally send only a portion of reported events to CloudWatch via the `:sample_rate` option. This parameter should contain a value between 0.0 and 1.0 (inclusive), and represents the proportion of events that will be reported. For instance, a sample rate value of 0.0 will ensure no events are sent, whereas a sample rate value of 1.0 will ensure all events are sent. A sample rate of 0.25 will result in roughly a quarter of all events being sent.
*Note:* Using a sampling rate of less than 1.0 will result in incorrect values for the "Sum" and "Counter" metric types (and some of the values like min/max for the "Summary" metric type) due to the under-reporting of events. You should only use a sampling rate of less than 1.0 if you have a guaranteed high rate of events and only care about summary statistics like averages and percentiles.
### Units
In order to report metrics in the CloudWatch UI, they must be one of the following values:
* Time units: `:second`, `:microsecond`, `:millisecond`
* Byte sizes: `:byte`, `:kilobyte`, `:megabyte`, `:gigabyte`, `:terabyte`
* Bit sizes: `:bit`, `:kilobit`, `:megabit`, `:gigabit`, `:terabit`
For `Telementry.Metrics.Counter`s, the unit will always be `:count`. Otherwise, the unit will be treated as `nil`.
### ExAws Setup
[`ExAws`](https://hexdocs.pm/ex_aws/ExAws.html) is the library used to send metrics to CloudWatch. Make sure your
[keys are configured](https://hexdocs.pm/ex_aws/ExAws.html#module-aws-key-configuration) and that they have the
[correct permissions](https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/permissions-reference-cw.html) of `cloudwatch:PutMetricData`.
Up to 10 tags are sent up to AWS as dimensions for a given metric.
## Running Tests
To run tests:
```shell
$ mix test
```
## Reporting Issues
Please report all issues [on github](https://github.com/bmuller/telemetry_metrics_cloudwatch/issues).
