skills.rest
erigontech
erigontechOFFICIAL
Explore all skills by this author →
WebsiteGitHub

benchmarkoor

Runs repeatable benchmarkoor performance benchmarks against Erigon using protected OverlayFS datadirs.

3.6k|1.5k|Updated May 27, 2019
Search Tags:#benchmarking#docker#Performance Testing#Ethereum#erigon#overlayfs#mgas/s
One-click install
npx skills add https://github.com/erigontech/erigon --skill benchmarkoor
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: benchmarkoor
Source: https://github.com/erigontech/erigon/tree/main/.claude/skills/benchmarkoor
Command: npx skills add https://github.com/erigontech/erigon --skill benchmarkoor

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Benchmarking an Ethereum client like Erigon against large State Actor datadirs or pre-populated snapshots risks mutating valuable pristine data, and ad-hoc runs produce results that cannot be compared across clients or commits. This Skill orchestrates benchmarkoor runs on Linux with kernel-enforced read-only OverlayFS protection so benchmarks are repeatable, safe, and auditable.

Core Features & Use Cases

  • Pristine datadir protection: Places a read-only bind mount in front of State Actor or pre-populated snapshots and routes all writes through disposable OverlayFS upper layers, with canary probes and fingerprint verification before and after runs.
  • Pinned, reproducible runs: Builds Erigon from a dedicated Git worktree, pins base images by digest, captures immutable image IDs, and records every repository commit for provenance.
  • Suite execution and validation: Discovers compute and stateful suite configs from benchmarkoor-tests, runs smoke subsets before full suites, and validates results against fixture inventories.
  • Cross-client ranking: Combines local results with peer runs from the Benchmarkoor website API to produce MGas/s comparison tables, rankings, and before/after improvement analysis.
  • Use Case: An Erigon developer benchmarking a local branch on a devnet can generate a pristine State Actor datadir, protect it with OverlayFS, run the full compute suite with a digest-pinned local image, and publish a PR table ranking the build against reth and other clients.

Quick Start

Ask the AI to run a benchmarkoor compute suite against the current Erigon worktree using a protected OverlayFS datadir and then produce a cross-client MGas/s ranking table.

Frequently Asked Questions about benchmarkoor

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I benchmark Erigon without modifying my datadir?▼

Place a kernel-enforced read-only bind mount in front of the pristine datadir and configure benchmarkoor with method: overlayfs so all writes go to a disposable upper layer. Verify the lower mount stays ro throughout the run and compare fingerprints before and after.

How do I run a benchmarkoor compute or stateful suite?▼

Resolve the suite's runner config from the benchmarkoor-tests checkout, build benchmarkoor from source, create a run-only override with your pinned image and protected source_dir, then run a 5-10 fixture smoke subset before the full suite. Validate results against the fixture inventory in the suite summary.

Can I benchmark a local Erigon branch instead of a released image?▼

Yes. Build the binary from a dedicated Git worktree, create a small overlay image copying that binary onto a digest-pinned base image, and reference the immutable image ID with pull_policy: never in the run override. Verify the image contains the byte-identical binary by comparing SHA-256 hashes.

How do I compare Erigon benchmark results against other clients?▼

Query the Benchmarkoor website API for peer runs matching your suite hash, extract per-fixture gas and time values, and compute MGas/s rates as gas_used times 1000 divided by time_ns. Rank clients per fixture and aggregate with weighted sums, never by averaging per-fixture rates.

Why must the pristine datadir never be used directly as a benchmark datadir?▼

Direct use risks mutating a valuable snapshot that took hours to build, and even reads through the writable path can update access times on relatime filesystems. The read-only bind plus OverlayFS upper keeps the original byte-identical, verified by metadata fingerprints and critical-file hashes.

What are the limitations of cross-host benchmark comparisons?▼

Cross-host or mixed datadir-method rankings measure the observed environments as a whole and cannot isolate client-code performance. Label local OverlayFS versus remote Schelk comparisons as directional rather than controlled measurements, and disclose every hardware and configuration mismatch.

Related Skills

Explore More Software Engineering Skills→
erigontecherigontech
3.6k

erigon-seg-retire

Automate Erigon segment retirement workflows to prepare publishable snapshots.

Official
Advanced
olaafrossiolaafrossi
10

benchmark

Benchmark web pages against baselines and report performance regressions.

Community
Advanced
jmagarjmagar
69

glances

Query Glances REST API for read-only CPU, memory, disk, and network metrics.

Community
Intermediate
StarryCodStarryCod
11

evaluating-llms-harness

Benchmark LLMs across 60+ tasks using lm-evaluation-harness backends.

Community
Advanced