I am first time user of Rally tool for benchmarking. Our QA cluster and Prod cluster was already built and we have indices in them.
1)Rally configuration guide states it downloads elastic and runs a track. How does this help in an existing cluster ? Is there a way to skip downloading ?
2)How does running an existing track help understanding the performance for my own indices ?
Hi,
A track is a specification of a benchmarking workload. You can either use the tracks that come out of the box with Rally or you can create your own tracks (which is recommended so you get an idea about the performance characteristics in a scenario that is realistic to you). I suggest you checkout the following sections in our docs:
Daniel
Hello Daniel,
Thank you. I would go through in detail on the provided links. On a surface level, it did not seem like there is an option to prevent elastic server being downloaded
Regards,
Narendran
Hi,
I misunderstood your original question and thought you wanted to avoid downloading the track but instead you want to avoid downloading Elasticsearch. You can achieve this by specifying --pipeline=benchmark-only and pointing Rally to your already running cluster. See our docs for details on this use case.
Daniel
Thank you Daniel. That was helpful information
I wonder what these warnings mean as it says the results could be misleading
"indicating that the cluster is not in a defined clean state."
Hi,
Rally does not control the cluster so you need to take extra care yourself that you benchmark under reproducible conditions (e.g. remove all persistent data, restart all nodes after each benchmark, etc.). For details see our blog post Seven Tips for Better Elasticsearch Benchmarks, specifically tip 6.
Daniel
This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.