Rally 0.5.0 released

Today we are proud to announce Rally 0.5.0 which brings a lot of improvements and fixes.

How to...?

  • ... upgrade: pip3 install --upgrade esrally
  • ... install: pip3 install esrally

Note: Depending on your system setup you probably need to prepend these commands with sudo.

Please follow the quickstart for a first time install though.


This release contains the following highlights:

  • Rally can now provision Elasticsearch clusters remotely (#184). For the moment, this is limited to single-machine "clusters" but we will expand remote provisioning to multiple machines next. From now on, this is the recommended mode in which you should run your benchmarks and our docs provide a complete example on how to run a remote benchmark.
  • Sparked by an initiative from @ebuildy, we have created a new repository at https://github.com/elastic/rally-results which allows us all to share results from benchmark that have been run on different hardware (think: bare metal, different EC2 instance types, Azure, GCE, ...). Rally can help you to gather the necessary meta-data by running esrally list facts --target-host=IP_OF_YOUR_TARGET_HOST. Over the next few days and weeks we will also transfer our own benchmarking results to this repo.
  • You can now use a new "nested" track that has been contributed by @Mark_Harwood which let's you test the performance of nested documents.
  • If you write your own track plugins, you can now use --enable-driver-profiling to check whether your track plugin has introduced a bottleneck.
  • Thanks to continued feedback from @Christian_Dahlqvist who pushes the functionality provided by Rally to its limits with his tracks, Rally provides now better support to structure your tracks.

The changelog contains the complete list of changes.

All users are recommended to upgrade to this version.

I want to thank everbody that raised issues, questions, provided other feedback or even contributed code and documentation. If you also want to get involved in Rally development, we have tagged a number of Github tickets which we think are easy to get started with. If you want to give it a try, just ask questions on the respective ticket.

What's next?

After we can benchmark single-machine "clusters" remotely, we will focus on expanding remote provisioning capabilities to multi-node clusters and also make benchmark candidate configuration more flexible.

Questions and Feedback

If you have questions or feedback, please just post in the Rally forum.