feat(scripts): include ribasushi-rpc-checks

.github/EXTERNAL_RPC_CHECKS_ISSUE_TEMPLATE.md

@@ -0,0 +1,8 @@
+---
+title: "[automated] External dataset RPC checks failure"
+labels: ["Bug"]
+---
+
+## Description
+
+Latest RPC checks against the external `data.riba.plus` dataset failed. Please [check the logs]({{ env.WORKFLOW_URL }}) for more information.

.github/workflows/external-rpc-checks.yml

@@ -0,0 +1,40 @@
+name: data.riba.plus RPC checks
+permissions:
+  contents: read
+  issues: write
+
+concurrency:
+  group: ${{ github.workflow }}
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: "0 13 * * *" # Runs every day at 13:00 UTC
+
+jobs:
+  rpc-checks:
+    name: RPC checks against the data.riba.plus dataset
+    runs-on: warp-ubuntu-2404-x64-8x
+    defaults:
+      run:
+        working-directory: scripts/tests/external-rpc-checks
+    steps:
+      - name: Checkout Sources
+        uses: actions/checkout@v7
+      - name: Run the RPC checks
+        run: ./setup.sh
+      - name: Dump docker logs
+        if: always()
+        uses: jwalton/gh-docker-logs@v2
+      - name: Set WORKFLOW_URL
+        if: always()
+        run: |
+          export WORKFLOW_URL="${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY}/actions/runs/${GITHUB_RUN_ID}"
+          echo "${WORKFLOW_URL}"
+          echo "WORKFLOW_URL=${WORKFLOW_URL}" >> "$GITHUB_ENV"
+      - uses: JasonEtco/create-an-issue@v2
+        if: github.ref == 'refs/heads/main' && failure()
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          filename: .github/EXTERNAL_RPC_CHECKS_ISSUE_TEMPLATE.md

docs/docs/developers/guides/rpc_api_compatibility.md

@@ -50,6 +50,12 @@ Forest.
 Forest does not yet support mining and none of the mining-related RPC calls will
 be implemented in the foreseeable future.
 
+## External dataset checks
+
+Forest is also checked against the external `data.riba.plus` dataset of recorded
+RPC responses, see `.github/workflows/external-rpc-checks.yml` and
+`scripts/tests/external-rpc-checks`.
+
 ## Gateway
 
 The `lotus-gateway` executable is a reverse-proxy that sanitizes RPC calls

scripts/tests/external-rpc-checks/.env

@@ -0,0 +1,2 @@
+FOREST_RPC_PORT=2345
+FOREST_HEALTHZ_RPC_PORT=2346

scripts/tests/external-rpc-checks/docker-compose.yaml

@@ -0,0 +1,68 @@
+name: external-rpc-checks
+
+x-forest-service: &forest-service
+  image: ghcr.io/chainsafe/forest:edge-fat
+  volumes:
+    - node-data:/data
+  environment:
+    # keeps the database in the container's volume, shared across containers
+    - FOREST_PATH=/data
+  user: 0:0
+
+services:
+  # Resolves yesterday's snapshot, imports it and back-fills the chain index.
+  init:
+    <<: *forest-service
+    volumes:
+      - node-data:/data
+      - ./init.sh:/init.sh:ro
+    entrypoint: ["/bin/bash", "/init.sh"]
+
+  forest:
+    <<: *forest-service
+    environment:
+      - FOREST_PATH=/data
+      - FOREST_CHAIN_INDEXER_ENABLED=1
+    command:
+      - --chain=calibnet
+      - --encrypt-keystore=false
+      - --no-gc
+      - --rpc-address=0.0.0.0:${FOREST_RPC_PORT}
+      - --healthcheck-address=0.0.0.0:${FOREST_HEALTHZ_RPC_PORT}
+    healthcheck:
+      # `/livez` from inside the container: RPC answering and peers connected.
+      test:
+        [
+          "CMD",
+          "forest-cli",
+          "healthcheck",
+          "live",
+          "--healthcheck-port",
+          "${FOREST_HEALTHZ_RPC_PORT}",
+        ]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 5m
+      start_interval: 5s
+
+  # Probes the node for the epochs the checks are about to query.
+  verify:
+    image: ghcr.io/chainsafe/forest-rpc-checks:latest
+    depends_on:
+      forest:
+        condition: service_healthy
+    volumes:
+      - ./verify.rb:/verify.rb:ro
+    entrypoint: ["ruby", "/verify.rb"]
+    environment:
+      FOREST_RPC_URL: forest:${FOREST_RPC_PORT}/rpc/v1
+
+  # The image entrypoints into `bundle exec ruby check_rpc.rb`, so only the epochs are appended.
+  rpc-checks:
+    image: ghcr.io/chainsafe/forest-rpc-checks:latest
+    environment:
+      FOREST_RPC_URL: forest:${FOREST_RPC_PORT}/rpc/v1
+
+volumes:
+  node-data:

scripts/tests/external-rpc-checks/init.sh

@@ -0,0 +1,30 @@
+#!/usr/bin/env bash
+# Runs in the `init` service: resolves the snapshot to test against, imports it
+# and back-fills the chain index, all before the daemon starts.
+
+set -euo pipefail
+
+# The Forest image ships neither curl nor jq.
+apt-get update -qq
+apt-get install -y -qq --no-install-recommends curl jq
+
+# The dataset lags the chain, so yesterday's snapshot is the one to test against.
+day=$(date -u -d '1 day ago' +%F)
+url=$(curl --silent --show-error --fail --retry 3 --connect-timeout 10 --max-time 60 \
+  "https://forest-archive.chainsafe.dev/list/calibnet/latest-v2?format=json" |
+  jq --raw-output --arg day "${day}" '[.items[].url | select(contains("_" + $day + "_"))] | first')
+[[ ${url} == https* ]] || {
+  echo "no calibnet snapshot published for ${day}"
+  exit 1
+}
+
+# Snapshot names end in the epoch of their head tipset. ./setup.sh reads it back
+# to pick the range to check.
+epoch=${url##*_height_}
+epoch=${epoch%%.*}
+printf '%s\n' "${epoch}" > /data/snapshot-epoch
+
+forest --chain=calibnet --encrypt-keystore=false --import-snapshot="${url}" --halt-after-import
+
+# Indexes the 1000 epochs below the snapshot head, inclusive on both ends
+forest-tool index backfill --chain=calibnet --from="${epoch}" --to="$((epoch - 1000))"

scripts/tests/external-rpc-checks/setup.sh

@@ -0,0 +1,25 @@
+#!/usr/bin/env bash
+# Runs the RPC checks against the external data.riba.plus dataset.
+# Needs docker only; everything else happens in containers.
+
+set -euo pipefail
+
+PARENT_PATH=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd -P)
+pushd "${PARENT_PATH}"
+
+# This should not be needed in GH. It is useful for running locally.
+docker compose down --remove-orphans --volumes
+
+# Imports the snapshot and back-fills the index, recording the head epoch.
+docker compose run --rm init
+SNAPSHOT_EPOCH="$(docker compose run --rm --no-TTY --entrypoint cat init /data/snapshot-epoch)"
+START=$((SNAPSHOT_EPOCH - 1000))
+END=$((SNAPSHOT_EPOCH - 1))
+
+docker compose up --detach --wait forest
+
+docker compose run --rm verify "${START}" "${END}"
+
+docker compose run --rm rpc-checks "${START}" "${END}"
+
+popd

scripts/tests/external-rpc-checks/verify.rb

@@ -0,0 +1,26 @@
+# frozen_string_literal: true
+
+# Confirms the back-fill indexed the epochs the checks query. Runs in a container
+# so the node's RPC port never has to be published.
+
+require 'json'
+require 'net/http'
+
+url = URI("http://#{ENV.fetch('FOREST_RPC_URL')}")
+
+ARGV.each do |argument|
+  epoch = Integer(argument)
+  request = Net::HTTP::Post.new(url, 'Content-Type' => 'application/json')
+  request.body = {
+    jsonrpc: '2.0',
+    id: 1,
+    method: 'eth_getBlockByNumber',
+    params: ["0x#{epoch.to_s(16)}", false]
+  }.to_json
+
+  response = Net::HTTP.start(url.hostname, url.port) { |http| http.request(request) }
+  body = JSON.parse(response.body)
+  abort "epoch #{epoch} is not indexed: #{response.code} #{response.body}" unless body.dig('result', 'number')
+
+  puts "epoch #{epoch} is indexed"
+end

src/tool/offline_server/server.rs

@@ -199,7 +199,7 @@ pub async fn start_offline_server(
     }
 
     // Validate tipsets since the {height} EPOCH when `height >= 0`,
-    // or valiadte the last {-height} EPOCH(s) when `height < 0`
+    // or validate the last {-height} EPOCH(s) when `height < 0`
     let validate_until_epoch = if height > 0 {
         height
     } else {