Misunderstanding of metrics

(Sergei Petrov) #1


On the previous week I wrote to you about the problem with custom plugin. (ES custom plugin testing).

As you suggested me, I created a separate ES cluster with my plugin and test data. Then I created a custom track for Rally and ran it. After a long time, I finally got the test result. But I don't understand why I got so huge numbers... For example: 99th percentile latency: 41007.5 ms? and the 99th percentile service_time: 56.3849 ms? Could you give more information about this metrics? Because I don't think that my TCP connection takes 41 secnods.

Some details about the query which I am testing: I create a query which used my plugin. Inside my plugin I use a TCP connection to the application and get the information which I need to the ES. If I executed the query from Kibana, I got the 5073 ms. It is time taken to receive the response. This time doesn't look like one of the metrics. Where I can get the time which I spend to receive the response from the ES similar to the Kibana?

Maybe I missed something, so could you explain to me why these numbers are so different? or why this could happen?

(Daniel Mitterdorfer) #2


you can find the answer in the FAQ: What does latency and service_time mean?

I hope that helps.


(system) #3

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.