running-benchmarks

name	running-benchmarks
description	Run Rally benchmarks (races) against Elasticsearch — an existing/external cluster or a Rally-provisioned distribution — and read the summary report. Use when running a race (any pipeline, track, challenge, target-hosts, or auth) or when interpreting throughput, latency, and service_time results.

Running Rally benchmarks

Launch a race against Elasticsearch and interpret its summary report. Rally can run benchmarks several ways (see Choose a pipeline).

For meaningful results, prefer an existing/external cluster you control: running Rally locally co-locates the load driver and the system under test on one host, where they can perturb each other's measurements. Local deployment is a good fit when that's not a concern — smoke testing, fast iterative local development, or demos.

Safety

• Never run Rally as root — Elasticsearch refuses to start with root privileges.
• Strongly avoid benchmarking production. Rally mutates cluster state (creating, writing to, and deleting indices), and competing traffic can perturb results. Prefer a dedicated, quiet environment.

Choose a pipeline

Rally infers the pipeline from your flags.

Pipeline	Use when	Selected by
benchmark-only	Primary path. Cluster is already running and you provisioned it.	--pipeline=benchmark-only --target-hosts=...
from-distribution	Rally downloads and runs Elasticsearch locally. Setup checks/demos; sharing a host skews results, so only large differences stand out.	--distribution-version=X
from-sources