Files
rustfs/.github/workflows/performance-ab.yml
T
Zhengchao An d73c8a783a ci(perf): cache the main baseline binary to cut the nightly double build (#4816)
Every push to main now builds the release binary once and stores it in the
actions cache as rustfs-baseline-<sha> (build-baseline-cache job). The A/B job
restores it for origin/main and passes --baseline-bin; on the nightly, where
the candidate commit equals the baseline, one cached binary serves both phases
with --skip-build and the run does zero source builds. A cache miss falls back
to the source double-build via --baseline-ref origin/main.

This removes the ~32min-per-side double build that pushed the 24-cell nightly
past its 120min ceiling (2026-07-11..07-14 all cancelled on timeout).

Also:
- alert-on-failure now fires on cancelled/timed-out, not just failure, so a
  timed-out nightly is no longer silent (the composite action already reports
  cancelled jobs); removed the stale perf-2 TODO now that the alert job exists.
- run_hotpath_warp_ab.sh appends a Provenance section to gate.md (baseline and
  candidate SHAs + binary source, runner, warp version, matrix params) via a
  repeatable --provenance-note; the gate exit code is preserved.

Refs rustfs/backlog#1152 (perf-3).
2026-07-15 09:33:56 +08:00

346 lines
15 KiB
YAML

# Copyright 2024 RustFS Team
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Warp A/B budget gate for the hotpath series (rustfs/backlog#935 HP-14, item 4).
#
# Two entry points, honestly scoped:
# * schedule (nightly, on main): post-merge detection — catches a regression
# within 24h of landing, not before merge.
# * pull_request labeled `perf-ab`: opt-in pre-merge gate for a specific PR.
# The `perf-deliberate-tradeoff` label runs the gate with --allow-regression so
# a deliberate correctness cost (e.g. the #4221 fsync durability fix) is
# recorded but does not block (rustfs/backlog#935 correction 1).
name: Performance A/B
on:
schedule:
- cron: "0 6 * * *" # 06:00 UTC nightly, against main
workflow_dispatch:
inputs:
duration:
description: "warp duration per round (short by default to fit the double-build budget)"
required: false
default: "12s"
type: string
allow_regression:
description: "Pass the gate despite a FAIL (deliberate tradeoff)"
required: false
default: false
type: boolean
pull_request:
types: [labeled, synchronize, reopened]
push:
# Every main commit pre-builds and caches its release binary (perf-3) so the
# nightly A/B restores a ready baseline instead of paying the double build.
branches: [main]
permissions:
contents: read
pull-requests: write
# Per-PR: a new push cancels the previous (up to 90-minute) A/B run instead of
# stacking them. Nightly schedule and manual dispatch get a unique group and
# always run to completion.
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.run_id }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
env:
CARGO_TERM_COLOR: always
RUST_BACKTRACE: 1
jobs:
# perf-3: on every push to main, build the release binary once and cache it
# keyed by commit SHA (rustfs-baseline-<sha>). The nightly A/B (and, later, the
# perf-7 PR gate) restore this instead of paying the ~32min-per-side source
# build. That double build is what pushed the expanded 24-cell nightly past its
# ceiling — 2026-07-11..07-14 all cancelled on the 120min timeout. Incremental
# builds off the shared cargo cache keep each push cheap, and building on the
# same sm-standard-2 runner the A/B measures on guarantees the cached binary is
# ABI-identical. Do NOT source this from build.yml's per-merge artifact: those
# are cancelled ~7/8 of the time and are not a reliable baseline.
build-baseline-cache:
name: Build + cache baseline binary
if: github.event_name == 'push'
runs-on: sm-standard-2
timeout-minutes: 60
steps:
- name: Checkout repository
uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
- name: Setup Rust environment
uses: ./.github/actions/setup
with:
rust-version: stable
cache-shared-key: warp-ab-${{ hashFiles('**/Cargo.lock') }}
cache-save-if: ${{ github.ref == 'refs/heads/main' }}
github-token: ${{ secrets.GITHUB_TOKEN }}
- name: Build release rustfs
run: cargo build --release --bin rustfs
- name: Stage binary for cache
run: |
set -euo pipefail
mkdir -p baseline-bin
cp target/release/rustfs baseline-bin/rustfs
- name: Cache baseline binary by SHA
uses: actions/cache/save@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6
with:
path: baseline-bin/rustfs
key: rustfs-baseline-${{ github.sha }}
warp-ab:
name: Warp A/B budget gate
# Always run on schedule / manual dispatch. Opt-in on PRs: only when the
# `perf-ab` label is present, and for `labeled` events only when the label
# being added is `perf-ab` itself (adding an unrelated label to an opted-in
# PR must not re-run the gate). Never on push — that event only feeds
# build-baseline-cache above.
if: >-
github.event_name == 'schedule' ||
github.event_name == 'workflow_dispatch' ||
(github.event_name == 'pull_request' &&
contains(github.event.pull_request.labels.*.name, 'perf-ab') &&
(github.event.action != 'labeled' || github.event.label.name == 'perf-ab'))
runs-on: sm-standard-2
# With perf-3's cached baseline binary the common (cache-hit) nightly is
# measurement-only and finishes well under 50min. This ceiling stays
# generous only to absorb the rare cache-miss fallback, which reverts to the
# ~65min source double-build; a timeout there now surfaces via the
# alert-on-failure job (it fires on cancelled/timed-out, not just failure).
# perf-6 recalibrates the budget once the noise study lands.
timeout-minutes: 120
steps:
- name: Checkout repository
uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
with:
fetch-depth: 0 # baseline is built from origin/main
- name: Setup Rust environment
uses: ./.github/actions/setup
with:
rust-version: stable
cache-shared-key: warp-ab-${{ hashFiles('**/Cargo.lock') }}
cache-save-if: ${{ github.ref == 'refs/heads/main' }}
github-token: ${{ secrets.GITHUB_TOKEN }}
- name: Install warp
run: |
set -euo pipefail
WARP_VERSION="1.0.0"
curl -fsSL "https://github.com/minio/warp/releases/download/v${WARP_VERSION}/warp_Linux_x86_64.tar.gz" \
| sudo tar -xz -C /usr/local/bin warp
warp --version
- name: Decide exemption
id: exempt
run: |
allow="false"
if [[ "${{ github.event_name }}" == "pull_request" ]] \
&& ${{ contains(github.event.pull_request.labels.*.name, 'perf-deliberate-tradeoff') }}; then
allow="true"
fi
if [[ "${{ github.event.inputs.allow_regression }}" == "true" ]]; then
allow="true"
fi
echo "allow_regression=$allow" >> "$GITHUB_OUTPUT"
# perf-3: resolve the commits so the cache can be keyed by SHA. The
# baseline is origin/main; the candidate is the checked-out ref. On the
# nightly (checkout == main) they are the same commit, so one cached binary
# serves both phases and the run does zero source builds.
- name: Resolve baseline / candidate commits
id: commits
run: |
set -euo pipefail
baseline_sha="$(git rev-parse origin/main)"
candidate_sha="$(git rev-parse HEAD)"
echo "baseline_sha=$baseline_sha" >> "$GITHUB_OUTPUT"
echo "candidate_sha=$candidate_sha" >> "$GITHUB_OUTPUT"
echo "baseline commit: $baseline_sha"
echo "candidate commit: $candidate_sha"
# Exact-key restore of the baseline binary built by build-baseline-cache
# when origin/main last landed. A miss (binary evicted or not built yet)
# leaves cache-hit unset and the rig falls back to a source build.
- name: Restore cached baseline binary
id: baseline_cache
uses: actions/cache/restore@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6
with:
path: baseline-bin/rustfs
key: rustfs-baseline-${{ steps.commits.outputs.baseline_sha }}
- name: Run warp A/B and gate
id: ab
run: |
set -euo pipefail
# Budget note: with perf-3's cached baseline the nightly does no source
# build on a cache hit, so the wall-clock is dominated by the short warp
# matrix — duration/rounds/cooldown are kept small to fit all 24 cells
# (6 workloads x 2 phases x 2 drive-sync) rather than dropping cells.
# --health-timeout 180 outlasts the server's own 120s startup-readiness
# budget, which the rig's previous 60s health poll undershot (the first
# two nightly failures). perf-6 recalibrates these once the noise study
# lands.
duration="${{ github.event.inputs.duration || '12s' }}"
baseline_sha="${{ steps.commits.outputs.baseline_sha }}"
candidate_sha="${{ steps.commits.outputs.candidate_sha }}"
baseline_hit="${{ steps.baseline_cache.outputs.cache-hit }}"
args=(--duration "$duration" --rounds 2 --cooldown 5 --health-timeout 180)
if [[ "$baseline_hit" == "true" ]]; then
chmod +x baseline-bin/rustfs
base_bin="$PWD/baseline-bin/rustfs"
args+=(--baseline-bin "$base_bin")
base_src="actions-cache (rustfs-baseline-$baseline_sha)"
if [[ "$candidate_sha" == "$baseline_sha" ]]; then
# Nightly on main: the candidate is the same commit as the baseline,
# so reuse the one cached binary for both phases and skip all builds.
args+=(--candidate-bin "$base_bin" --skip-build)
cand_src="actions-cache (same commit as baseline)"
else
cand_src="source build of the checked-out ref"
fi
else
# Cache miss (binary evicted or not built for this SHA): fall back to
# the slow source double-build so the run still produces a result.
args+=(--baseline-ref origin/main)
base_src="source build of origin/main (cache miss)"
cand_src="source build of the checked-out ref"
fi
args+=(--provenance-note "baseline commit: $baseline_sha - $base_src")
args+=(--provenance-note "candidate commit: $candidate_sha - $cand_src")
if [[ "${{ steps.exempt.outputs.allow_regression }}" == "true" ]]; then
args+=(--allow-regression --exemption-reason "labeled perf-deliberate-tradeoff / dispatch override")
fi
# Do not let a gate FAIL abort the job here; capture status and surface
# it after the PR comment is posted.
set +e
bash scripts/run_hotpath_warp_ab.sh "${args[@]}"
echo "status=$?" >> "$GITHUB_OUTPUT"
set -e
# Locate the newest run dir + gate.md for the summary/comment/artifact
# steps. On a startup failure there is no gate.md, but the run dir still
# holds server-logs/ for diagnosis.
# Run dirs are UTC-timestamp names (no special chars); ls is safe here.
# shellcheck disable=SC2012
run_dir="$(ls -td target/hotpath-ab/*/ 2>/dev/null | head -n1 || true)"
echo "run_dir=${run_dir%/}" >> "$GITHUB_OUTPUT"
# shellcheck disable=SC2012
gate_md="$(ls -t target/hotpath-ab/*/gate.md 2>/dev/null | head -n1 || true)"
echo "gate_md=$gate_md" >> "$GITHUB_OUTPUT"
- name: Upload A/B results
if: always()
uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6
with:
name: hotpath-warp-ab-${{ github.run_number }}
# Includes per-cell median_summary.csv / baseline_compare.csv, gate.md,
# and server-logs/ (rustfs.log + startup env per phase) so a failed run
# is diagnosable. Short retention: this is churny nightly debug data.
path: target/hotpath-ab/
if-no-files-found: warn
retention-days: 14
- name: Write gate summary
if: always()
run: |
set -euo pipefail
status="${{ steps.ab.outputs.status }}"
gate_md="${{ steps.ab.outputs.gate_md }}"
run_dir="${{ steps.ab.outputs.run_dir }}"
{
echo "## Hotpath warp A/B — run ${{ github.run_number }}"
echo
if [[ "$status" == "0" ]]; then
echo "Rig/gate exit: \`0\` (pass or warn)."
else
echo "Rig/gate exit: \`${status:-unknown}\` — **FAILED**."
fi
echo
if [[ -n "$gate_md" && -f "$gate_md" ]]; then
cat "$gate_md"
else
echo "No \`gate.md\` produced — the rig failed **before** the gate"
echo "(most likely server startup / health). Failing phase(s) below;"
echo "full logs in the \`hotpath-warp-ab-${{ github.run_number }}\` artifact."
if [[ -n "$run_dir" && -d "$run_dir/server-logs" ]]; then
for f in "$run_dir"/server-logs/*.log; do
[[ -f "$f" ]] || continue
echo
echo "<details><summary>$(basename "$f")</summary>"
echo
echo '```'
tail -n 30 "$f"
echo '```'
echo "</details>"
done
fi
fi
} >> "$GITHUB_STEP_SUMMARY"
- name: Comment gate result on PR
if: always() && github.event_name == 'pull_request' && steps.ab.outputs.gate_md != ''
env:
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
run: |
gh pr comment "${{ github.event.pull_request.number }}" --body-file "${{ steps.ab.outputs.gate_md }}"
# Scheduled failure alerting is handled by the alert-on-failure job below
# (perf-2 consuming ci-8's schedule-failure-issue composite action).
- name: Enforce gate
if: always()
run: |
status="${{ steps.ab.outputs.status }}"
if [[ "$status" != "0" ]]; then
echo "::error::warp A/B budget gate failed (exit $status). See the step summary / PR comment / gate.md artifact." >&2
exit "$status"
fi
echo "warp A/B budget gate passed."
alert-on-failure:
name: Alert on scheduled failure
needs: [warp-ab]
# `always()` is required: without it this job is skipped when a needed
# job fails. Alerts only for scheduled (nightly) runs (backlog#1149
# ci-8); PR and manual dispatch failures are already watched by a human.
# `cancelled` is included alongside `failure` on purpose: a job that hits
# timeout-minutes ends as `cancelled`, and the 2026-07-11..07-14 nightly
# timeouts went silent precisely because the guard was failure-only. The
# composite action already reports cancelled/timed-out jobs in the issue
# body. (Scheduled runs get a unique concurrency group with
# cancel-in-progress off, so a cancellation here means a timeout/manual
# abort, never a superseding run.)
if: >-
always() && github.event_name == 'schedule' &&
(contains(needs.*.result, 'failure') || contains(needs.*.result, 'cancelled'))
runs-on: ubuntu-latest
timeout-minutes: 10
permissions:
contents: read
issues: write
steps:
- uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
- name: Open or update failure-tracking issue
uses: ./.github/actions/schedule-failure-issue
with:
github-token: ${{ secrets.GITHUB_TOKEN }}