How should I filter multiple fields with the same name in logstash?

I'm putting tsung logs into ElasticSearch (ES) so that I can filter, visualize and compare results using Kibana.

I'm using logstash and its JSON parsing filter to push tsung logs in JSON format to ES.

Tsung logs are a bit complicated (IMO) with array objects into array objects, multiple-lines event, and several fields having the same name such as "value" in my example hereafter.

I would like to transform this event:

{
 "stats":[
  {"timestamp": 1317413861,  "samples": [
   {"name": "users", "value": 0, "max": 1},
   {"name": "users_count", "value": 1, "total": 1},
   {"name": "finish_users_count", "value": 1, "total": 1}]}]}

into this:

{"timestamp": 1317413861},{"users_value":0},{"users_max":1},{"users_count_value":1},{"users_count_total":1},{"finish_users_count_value":1},{"finish_users_count_total":1}

Since the entire tsung log file is forwarded to logstash at the end of a performance test campaign, I'm thinking about using regex to remove CR and unusefull stats and samples arrays before sending the event to logstash in order to simplify a little bit.

And then, I would use those kind of JSON filter options:

add_field => {"%{name}_value" => "%{value}"}
add_field => {"%{name}_max" => "%{max}"}
add_field => {"%{name}_total" => "%{total}"}

But how should I handle the fact that there are many value fields in one event for instance? What is the best thing to do?

Thanks for your help.

Solution

Feels like the ruby{} filter would be needed here. Loop across the entries in the 'samples' field, and construct your own fields based on the name/value/total/max.

There are examples of this type of behavior elsewhere on SO.