(data)dogdirect

Directly send metrics to DataDog via the HTTP API using golang.

What Problem Are We Solving?

Using DataDog's SasS offering to upload time series data when you can't use the official DataDog agent.

In general, you should use the official DataDog agent. It's great. Use it. But, sometimes you can't or won't run the official DataDog agent. This could be because:

You are on a constrained environment with low resources
You don't want alien code running in your environment
You don't want to run agent process along side your application.
You don't want to run dedicated container to run the DataDog agent (per cluster or per availability zone).

So if you can't (or won't run the datadog), package provides a simple interface for counters and gauges and uploads data using the HTTP API (the same one the DataDog agent uses).

What does this do?

Stores your metrics locally supporting both counters and gauges.
Timinig and Histograms are supported but the descriptive statistics are hardwired and not configurable (but easy to add).
Allows setting a global namespace
Allows setting global tags (applied to every metric)
Uploads your metrics to DataDog every 15 seconds

What doesn't this do?

Anything that's not a guage or counter or histogram: logs, traces, service checks, events.
Error handling is probably not awesome. Pull requests welcome.

This is not an official client

This is not an official client. This project is not sponsored or blessed by DataDog.

References and Credits

Offical HTTP API Documentation

THe offical HTTP API documentation on metrics

We store metric points at the 1 second resolution, but we’d prefer if you only submitted points every 15 seconds. Any metrics with fractions of a second timestamps gets rounded to the nearest second, and if any points have the same timestamp, the latest point overwrites the previous ones.

https://docs.datadoghq.com/api/?lang=bash#post-timeseries-points

datadog-go

In particular, much of this code is based on statsd.go. I have mixed feelings about the "watcher" being glued into the main object but seems to work for now.

Interestingly, the buffered implimentation doesn't consolidate anything. If you do 100 increments of a single stat, it will send 100 statsd messages.

datadog-agent

The golang datadog agent has some nuggest, especially the metrics. Unfortunately the metrics package can't be used standalone as it glues in a bunch of extraneous stuff specific to DataDog (logging, configuration). That's too bad.

strip/veneue

stripe/veneue is a statds server on sterioids. It uses the HTTP API as well.

See the datadog section and the http.go for details.

signalsciences / dogdirect