Almost two years after the release of Rally 1.0.0 we are excited to announce Rally 2.0.0 today as a collaboration from 40 contributors. Many thanks to all our community contributors for the numerous discussions, bug reports and pull requests!
This release contains several improvements and bug fixes. We’d like to highlight some of them below:
Rally helps you to create benchmarks from your cluster’s data
Previously you needed to extract all data yourself when creating a new Rally track. With Rally 2.0.0 Rally can create a track based on your cluster’s data automatically for you with a command as simple as:
esrally create-track --track=acme --target-hosts=127.0.0.1:9200 --indices="products,companies" --output-path=~/tracks
For more details check our track tutorial.
New load generator
So far Rally has been optimized for benchmarks with a small number of clients. With Rally 2.0.0 we have reimplemented the load generator so that it can simulate significantly more concurrent clients by running them in an event loop.
Faster decompression of document corpora
As one of the first steps in a benchmark, Rally needs to decompress the document corpora which are often several GB large. So far, Rally has only used one CPU core for decompression but with 2.0.0 Rally uses all available CPU cores for decompression if pbzip2 is available on the PATH speeding up decompression significantly.
Before upgrading from an earlier version of Rally, check our migration guide. Most notably, Rally 2.0.0 requires at least Python 3.8.0.
See the 2.0.0 release page for all details.
While Rally helps you to create proper benchmarks there are also a lot of methodological aspects to consider so be sure to check our blog post that provides seven tips for better Elasticsearch benchmarks.
How to install Rally?
See our installation guide.
Questions and Feedback
If you have questions or feedback, post in the Rally forum.