README.md

Etsy's Logstash Plugins

This repo contains a number of logstash plugins used at Etsy.

hashid

filter {
  hashid { }
}

hashid generates a SHA1 hash on the message field, and uses that to set the _id used by Elasticsearch. This makes it possible to reply whole logs and avoid duplicate entires in Elasticsearch. It comes at a cost of efficiency in Elasticsearch. Elasticsearch by default uses fluke UUIDs to make make searching for results more predictable. Using truly random IDs removes this predictability. I many cases the hit at search time isn't terrible and hardly noticeable, with quite a positive upside for log replay.

derrick

filter {
  derrick { }
}

derrick parses the http_headers field and looks for data generated by Derrick to decode. derrick is a handy packet capturing utility which combines multiple packets from a TCP stream and prints them as a single line to a file. You can call derrick like this to capture traffic:

/usr/bin/derrick -i eth0 -m -f "port 80" > /tmp/derrick.log

Then ship those logs to logstash.

splitkv

filter {
  splitkv {
    field_split => "&"
    value_split => "="
  }
}

In the case of very well formed fields which need to be split on a conistent field separator, and a consistent key/value separator, throughout the field, splitkv offers significant performance improvements over the default kv plugin. It only works on fields where the entire set of data can be split consistently. An example of this would be an HTTP query string:

foo1=bar1&foo2=bar2&food3=bar3&foo4=bar4

The plugin first calls split('&') on the whole string, followed by split('=') on each of the resulting elements. By default, the field separator is a blank space, and the key/value separator is =.