Benchmark our own Cluster and Data

[sancroth (Nick)]

I would like to run some benchmarks against our own machines and data.

This means that i don't want to create new indices etc. I only wish to run search queries and benchmark the response times.

We are using elasticsearch for an e-shop with a huge catalog and i wish to see the behavior of searches under a stressful period. Pretty much throw the team on it while a bench is running and simulate a loaded period.

I have found that i can give the hosts i want to run my tracks on. But i am not sure how to target my indices and where to pass the queries to be run etc.

Is the above scenario feasible or an index needs to be created first and then the tool matches the data given against the data indexed as search queries?

I've already given a read on https://esrally.readthedocs.io/en/stable/recipes.html#changing-the-default-track-repository but other than the how to bench your host, there is no info on how to use your own indices and queries.

[dliappis (Dimitrios Liappis)]

Hi @sancroth,

What you want is possible but you will need to create a custom track for your use case (which is anyway the best practice for benchmarking individual cases).

You can take a look at the track reference examples where you see a minimal track just executing one query against existing indices.

Dimitris