linera validator benchmark

A pre-onboarding benchmark for a single candidate validator: probe its serving capacity, network path, and (optionally) block-ingest throughput before adding it to the committee. It targets one validator, emits a structured report, and gives no verdict — the operator reads the numbers and decides.

Quick start

# Candidate that already follows the chain(s) you pass:
linera validator benchmark grpcs:host:443 \
  --chain <CHAIN_ID> \
  --output md

# Cold candidate (no blocks yet): seed first, then the read layers run on real data:
linera validator benchmark grpcs:host:443 \
  --chain <CHAIN_ID> --deep \
  --output "json:report.json,md:report.md,brief"

Address format is grpcs:host:port (scheme, host, port separated by single colons — not a :// URL).

What it runs

In execution order: L1 preflight · L2 partial sync (seed, opt-in --deep) · L3 read baseline · L4 read stress · L5 bulk download · L6 tip-lag.

The read layers (L3–L5) are only meaningful if the candidate already holds the --chain. A not-yet-committee candidate may hold nothing — pass --deep to seed blocks first, or pre-sync it (linera validator sync). The tool warns when a chain is not held.

Output

--output <format>[:<path>], repeatable or ,/+-separated. Formats: json, yaml, md, brief (default md to stdout). Files are rewritten after each layer, so an interrupt still leaves the completed layers on disk.

More

• REFERENCE.md — full reference: layer semantics, report schema, and all options.
• examples/ — a real run captured in every format.
• The authoritative, always-current flag list is generated from the code into the repo's top-level CLI.md (linera validator help-markdown).