Arithmetic between count queries in histograms

I have some Logstash events with 2 fields:

• a field indicating a success or error
• a field indicating the Autonomous System number of the event origin

I can currently/easily draw on the same histogram the counts of error event for, say, 5 selected AS.
Unfortunately it has a semantic drawback: if an AS is overly represented, it will probably also have more error events, even if that AS is no more faulty than another in percentage.

So I'd like to draw, on the same histogram, for, the same 5 AS, the percentage of errors by doing, for each bucket, count(errors(AS#n)) * 100 / (count(errors(AS#n)) + count(success(AS#n))).

Would it be doable with Kibana? Should I do some filtering in Logstash? Or should I forward some metrics to Graphite/statsd and continue from there?

[msupino]

any update on this ? i need something similar.. cant find examples

[rashidkpc]

This would really be better accomplished in Graphite. We have no immediate plans to implement time series math in Kibana.