About StatsD

StatsD is a tool developed by Etsy and Flickr (complicated story).
Its main use is providing a middleman for Graphite, which is a real-time graphing tool, mainly used for performance metrics.
Thanks to StatsD, I can use TCP sockets between servers (guaranteeing packet delivery), while processes generating the data-points are using a localhost UDP socket (minimal overhead without losing packets), freeing them to do some real work.
Sometimes, things don’t work.

Basic Troubleshooting

I don’t claim to be an expert on StatsD, but I didn’t find a good post about troubleshooting, so I thought I’ll list some of my methods. I assume the problem we’re investigating is the most common and elusive of all: Data sent to StatsD isn’t showing up on Graphite

Send some Spam

Try sending some data to both StatsD and Graphite and see what sticks:


# Send to Graphite
echo "test.back.slash 4 `date +%s`" | nc -C ${GRAPHITE_SERVER} ${GRAPHITE_CARBON_PORT}

# Test
curl "http://${GRAPHITE_SERVER}:${GRAPHITE_WEBUI_PORT}/render/?target=${TEST_STRING}&from=-60sec&format=json"

# Send to StatsD
echo 'test.back.slash:1|c' | nc -C -w1 -u ${STATSD_SERVER} ${STATSD_PORT}

# Test
curl "http://${GRAPHITE_SERVER}:${GRAPHITE_WEBUI_PORT}/render/?target=stats.gauges.${TEST_STRING}&from=-60sec&format=json"

Redirect to Console

First, modify your StatsD configuration file to add the console backend (something like backends: ["./backends/graphite", "./backends/console"]).
Then start StatsD (like node /usr/local/statsd/stats.js /your/config.js), and watch the output. Check if your data is there.


Modify the StatsD configuration file to contain debug: true, and see if you get some interesting output in stderr.


This might seem like a last-ditch effort, but you can always use tcpdump to record interesting traffic.
I use something like:

sudo tcpdump -i any -w statsd-ts.dump "(port $GRAPHITE_CARBON_PORT) or (port $STATSD_PORT)"

Using this dump you can try and see if something’s fishy.
Since both StatsD and Graphite use packets that contain readable text, I use this combo to parse all textual data in ruby.
It’s not perfect, but it works.

`tshark -r #{filename} -T fields -e data.data`.split("\n").join(':').split(":").map{|m|m.hex.chr}.join.split("\n")[1..-2].map{|x|arr=x.split(' ');{'stat'=>arr[0],'val'=>arr[1],'time'=>arr[2].to_i}}