What is the risk of running Rally on production?

I have ran Rally on a staging instance and wanted to compare the performance to our existing production cluster. From the documentation:

First of all: Please (please, please) do NOT run Rally against your production cluster if you are just getting started with it. You have been warned.

If this is regarding creating and deleting indexes with the same names as the track, it won't be an issue. Will Rally affect the performance/take down the production cluster? I saw minimal latency issues when running against our staging instance.

Yes, it's about the 2 issues that were mentioned:

  • As part of a race Rally prepares cluster configuration which includes steps such as index and index template removal (typically prior to subsequent recreation but with content determined by a Rally track, not the content that could have been present in the production cluster).
  • Rally can generate enough traffic to affect performance of the target cluster.

The exact impact depends on a Rally track definition vs. what's currently being used in the target cluster, track parameters (e.g. number of clients) and load driver(s) specification when compared with target cluster resources.