mirror of
https://github.com/rustfs/rustfs.git
synced 2026-10-27 14:27:10 +00:00
ci: remove never-green Performance A/B workflow (#8979)
This commit is contained in:
hector
@@ -74,7 +74,6 @@ Notes:
|
||||
- `.github/workflows/e2e-s3tests.yml` (weekly full s3-tests sweep)
|
||||
- `.github/workflows/mint.yml` (weekly multi-SDK mint run)
|
||||
- `.github/workflows/fuzz.yml` (nightly fuzz corpus jobs)
|
||||
- `.github/workflows/performance-ab.yml` (nightly warp A/B gate)
|
||||
- **Pending:** the ci-7 e2e nightly-full workflow does not exist yet
|
||||
(backlog#1149 ci-7). When it lands, wire it up with the snippet above.
|
||||
|
||||
|
||||
@@ -18,7 +18,6 @@
|
||||
"never_ran_grace_until": "2026-09-08T00:00:00Z"
|
||||
},
|
||||
{ "workflow": ".github/workflows/nightly-gnu.yml", "max_age_hours": 36 },
|
||||
{ "workflow": ".github/workflows/performance-ab.yml", "max_age_hours": 36 },
|
||||
{
|
||||
"workflow": ".github/workflows/runner-hygiene.yml",
|
||||
"max_age_hours": 792,
|
||||
|
||||
@@ -29,7 +29,6 @@ on:
|
||||
- 'scripts/test_package_service_scripts.sh'
|
||||
- 'scripts/test_package_versions.sh'
|
||||
- 'scripts/test_docker_workflow.py'
|
||||
- 'scripts/security/check_performance_ab_workflow.sh'
|
||||
- 'scripts/security/check_preview_release_workflow.sh'
|
||||
- 'scripts/security/check_tier_artifact_workflow.sh'
|
||||
- 'scripts/security/check_workflow_pins.sh'
|
||||
@@ -48,7 +47,6 @@ on:
|
||||
- 'scripts/test_package_service_scripts.sh'
|
||||
- 'scripts/test_package_versions.sh'
|
||||
- 'scripts/test_docker_workflow.py'
|
||||
- 'scripts/security/check_performance_ab_workflow.sh'
|
||||
- 'scripts/security/check_preview_release_workflow.sh'
|
||||
- 'scripts/security/check_tier_artifact_workflow.sh'
|
||||
- 'scripts/security/check_workflow_pins.sh'
|
||||
@@ -155,9 +153,6 @@ jobs:
|
||||
- name: Check preview release workflow policy
|
||||
run: ./scripts/security/check_preview_release_workflow.sh
|
||||
|
||||
- name: Check performance A/B workflow trust boundary
|
||||
run: ./scripts/security/check_performance_ab_workflow.sh
|
||||
|
||||
- name: Check tier evidence workflow isolation
|
||||
run: ./scripts/security/check_tier_artifact_workflow.sh
|
||||
|
||||
|
||||
@@ -45,7 +45,7 @@ on:
|
||||
# 07:00 UTC Sunday — staggered clear of the other Sunday crons: ci (00:00),
|
||||
# build (01:00), e2e-s3tests (02:00), audit (03:00), nix-flake-update
|
||||
# (05:00), mint (06:00), and the daily fuzz (02:00), minio-interop (03:17),
|
||||
# e2e-replication-nightly (04:00) and performance-ab (06:00) lanes.
|
||||
# and e2e-replication-nightly (04:00) lanes.
|
||||
- cron: "43 7 * * 0"
|
||||
|
||||
concurrency:
|
||||
|
||||
@@ -38,8 +38,8 @@ name: e2e-nightly
|
||||
on:
|
||||
workflow_dispatch:
|
||||
schedule:
|
||||
# 04:00 UTC nightly — staggered clear of fuzz/e2e-s3tests (02:00),
|
||||
# stale (01:30) and performance-ab (06:00).
|
||||
# 04:00 UTC nightly — staggered clear of fuzz/e2e-s3tests (02:00)
|
||||
# and stale (01:30).
|
||||
- cron: "29 4 * * *"
|
||||
|
||||
# Only alert-on-failure needs more than read access; it declares its own
|
||||
|
||||
@@ -1,384 +0,0 @@
|
||||
# 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.
|
||||
# * workflow_dispatch: an explicitly selected trusted ref.
|
||||
# The dispatch input can run the gate with --allow-regression so a deliberate
|
||||
# correctness cost (e.g. the #4221 fsync durability fix) is recorded, not
|
||||
# blocked (rustfs/backlog#935 correction 1).
|
||||
|
||||
name: Performance A/B
|
||||
|
||||
on:
|
||||
schedule:
|
||||
- cron: "31 6 * * *" # 06:31 UTC nightly, against main
|
||||
workflow_dispatch:
|
||||
inputs:
|
||||
duration:
|
||||
description: "warp duration per round"
|
||||
required: false
|
||||
default: "12s"
|
||||
type: string
|
||||
allow_regression:
|
||||
description: "Pass the gate despite a FAIL (deliberate tradeoff)"
|
||||
required: false
|
||||
default: false
|
||||
type: boolean
|
||||
permissions:
|
||||
actions: read
|
||||
contents: read
|
||||
|
||||
env:
|
||||
CARGO_TERM_COLOR: always
|
||||
RUST_BACKTRACE: 1
|
||||
|
||||
jobs:
|
||||
warp-ab:
|
||||
if: github.repository == 'rustfs/rustfs'
|
||||
name: Warp A/B budget gate
|
||||
runs-on: sm-standard-2
|
||||
# A normal nightly restores the last successful binary and builds only the
|
||||
# candidate; daily access keeps that cache warm. A cache miss may build both
|
||||
# and needs room for the A/B run plus artifact and cache publication.
|
||||
timeout-minutes: 180
|
||||
steps:
|
||||
- name: Checkout repository
|
||||
uses: actions/checkout@f548e57e544e1ff5a4c46bf1e1b8685f8e4a348a # v7
|
||||
with:
|
||||
persist-credentials: false
|
||||
fetch-depth: 0 # baseline may be an earlier successful scheduled head
|
||||
|
||||
- name: Setup Rust environment
|
||||
uses: ./.github/actions/setup
|
||||
with:
|
||||
rust-version: stable
|
||||
cache-shared-key: warp-ab-${{ hashFiles('**/Cargo.lock') }}
|
||||
cache-save-if: ${{ github.ref == 'refs/heads/main' }}
|
||||
|
||||
- name: Install warp
|
||||
run: |
|
||||
set -euo pipefail
|
||||
WARP_VERSION="1.0.0"
|
||||
curl -fsSL "https://github.com/minio/warp/releases/download/v${WARP_VERSION}/warp_Linux_x86_64.tar.gz" \
|
||||
| sudo tar -xz -C /usr/local/bin warp
|
||||
warp --version
|
||||
|
||||
- name: Decide exemption
|
||||
id: exempt
|
||||
env:
|
||||
INPUT_ALLOW_REGRESSION: ${{ github.event.inputs.allow_regression }}
|
||||
run: |
|
||||
allow="false"
|
||||
if [[ "$INPUT_ALLOW_REGRESSION" == "true" ]]; then
|
||||
allow="true"
|
||||
fi
|
||||
echo "allow_regression=$allow" >> "$GITHUB_OUTPUT"
|
||||
|
||||
# A failed regression run must keep comparing against the last known-good
|
||||
# scheduled head. Otherwise the next nightly would absorb the regression
|
||||
# into its baseline and turn green without a fix.
|
||||
- name: Find last successful scheduled baseline
|
||||
id: scheduled_baseline
|
||||
if: github.event_name == 'schedule'
|
||||
uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
|
||||
with:
|
||||
result-encoding: string
|
||||
script: |
|
||||
const { data } = await github.rest.actions.listWorkflowRuns({
|
||||
owner: context.repo.owner,
|
||||
repo: context.repo.repo,
|
||||
workflow_id: "performance-ab.yml",
|
||||
event: "schedule",
|
||||
status: "success",
|
||||
per_page: 1,
|
||||
});
|
||||
return data.workflow_runs[0]?.head_sha ?? "";
|
||||
|
||||
# Manual runs compare a selected ref with current main. Scheduled runs
|
||||
# compare current main with the last successful scheduled head. With no
|
||||
# history, the first run measures the candidate against itself and seeds
|
||||
# that head only if the complete rig succeeds.
|
||||
- name: Resolve baseline / candidate commits
|
||||
id: commits
|
||||
env:
|
||||
SCHEDULED_BASELINE_SHA: ${{ steps.scheduled_baseline.outputs.result }}
|
||||
run: |
|
||||
set -euo pipefail
|
||||
candidate_sha="$(git rev-parse HEAD)"
|
||||
if [[ "${{ github.event_name }}" == "schedule" ]]; then
|
||||
baseline_sha="${SCHEDULED_BASELINE_SHA:-$candidate_sha}"
|
||||
else
|
||||
baseline_sha="$(git rev-parse origin/main)"
|
||||
fi
|
||||
git cat-file -e "${baseline_sha}^{commit}"
|
||||
if ! git merge-base --is-ancestor "$baseline_sha" "$candidate_sha"; then
|
||||
echo "::error::baseline $baseline_sha is not an ancestor of candidate $candidate_sha; update the selected ref before comparing" >&2
|
||||
exit 1
|
||||
fi
|
||||
echo "baseline_sha=$baseline_sha" >> "$GITHUB_OUTPUT"
|
||||
echo "candidate_sha=$candidate_sha" >> "$GITHUB_OUTPUT"
|
||||
echo "baseline commit: $baseline_sha"
|
||||
echo "candidate commit: $candidate_sha"
|
||||
|
||||
# Exact-key restore of the candidate binary saved by its successful
|
||||
# scheduled run. A miss leaves cache-hit unset and falls back to a source
|
||||
# build of that known-good head.
|
||||
- name: Restore cached baseline binary
|
||||
id: baseline_cache
|
||||
uses: actions/cache/restore@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6
|
||||
with:
|
||||
path: baseline-bin/rustfs
|
||||
key: rustfs-baseline-${{ steps.commits.outputs.baseline_sha }}
|
||||
|
||||
# Self-heal: on a nightly/dispatch run where the candidate commit IS the
|
||||
# baseline commit, a cache miss would make the rig build the same commit
|
||||
# twice (~65min per side with the post-#4806 LTO profile — no job budget
|
||||
# fits that). Build it once here, reuse it for both phases, and save it
|
||||
# back to the cache so the next run hits.
|
||||
- name: Build baseline on cache miss (same-commit self-heal)
|
||||
id: selfheal
|
||||
if: >-
|
||||
steps.baseline_cache.outputs.cache-hit != 'true' &&
|
||||
steps.commits.outputs.baseline_sha == steps.commits.outputs.candidate_sha
|
||||
run: |
|
||||
set -euo pipefail
|
||||
cargo build --release --bin rustfs
|
||||
mkdir -p baseline-bin
|
||||
cp target/release/rustfs baseline-bin/rustfs
|
||||
echo "built=true" >> "$GITHUB_OUTPUT"
|
||||
|
||||
- name: Build baseline on cache miss (different candidate)
|
||||
id: baseline_build
|
||||
if: >-
|
||||
steps.baseline_cache.outputs.cache-hit != 'true' &&
|
||||
steps.commits.outputs.baseline_sha != steps.commits.outputs.candidate_sha
|
||||
run: |
|
||||
set -euo pipefail
|
||||
baseline_root="$RUNNER_TEMP/rustfs-baseline-${{ github.run_id }}"
|
||||
baseline_target="$RUNNER_TEMP/rustfs-baseline-target-${{ github.run_id }}"
|
||||
git worktree add --detach "$baseline_root" "${{ steps.commits.outputs.baseline_sha }}"
|
||||
cargo build --release --manifest-path "$baseline_root/Cargo.toml" --bin rustfs --target-dir "$baseline_target"
|
||||
mkdir -p baseline-bin
|
||||
cp "$baseline_target/release/rustfs" baseline-bin/rustfs
|
||||
git worktree remove --force "$baseline_root"
|
||||
echo "built=true" >> "$GITHUB_OUTPUT"
|
||||
|
||||
- name: Build candidate binary
|
||||
id: candidate_build
|
||||
if: steps.commits.outputs.baseline_sha != steps.commits.outputs.candidate_sha
|
||||
run: |
|
||||
set -euo pipefail
|
||||
cargo build --release --bin rustfs
|
||||
mkdir -p candidate-bin
|
||||
cp target/release/rustfs candidate-bin/rustfs
|
||||
echo "built=true" >> "$GITHUB_OUTPUT"
|
||||
|
||||
- name: Save self-healed baseline to cache
|
||||
if: steps.selfheal.outputs.built == 'true' || steps.baseline_build.outputs.built == 'true'
|
||||
uses: actions/cache/save@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6
|
||||
with:
|
||||
path: baseline-bin/rustfs
|
||||
key: rustfs-baseline-${{ steps.commits.outputs.baseline_sha }}
|
||||
|
||||
- name: Run warp A/B and gate
|
||||
id: ab
|
||||
env:
|
||||
INPUT_DURATION: ${{ github.event.inputs.duration }}
|
||||
run: |
|
||||
set -euo pipefail
|
||||
# The formal runner executes A1 baseline -> B1 candidate -> B2 candidate
|
||||
# -> A2 baseline for each workload and drive-sync cell. It requires three
|
||||
# rounds per leg to emit tail latency and error-rate evidence.
|
||||
# --health-timeout 180 outlasts the server's own 120s startup-readiness
|
||||
# budget, which the rig's previous 60s health poll undershot (the first
|
||||
# two nightly failures). perf-6 recalibrates these once the noise study
|
||||
# lands.
|
||||
duration="${INPUT_DURATION:-12s}"
|
||||
baseline_sha="${{ steps.commits.outputs.baseline_sha }}"
|
||||
candidate_sha="${{ steps.commits.outputs.candidate_sha }}"
|
||||
baseline_hit="${{ steps.baseline_cache.outputs.cache-hit }}"
|
||||
selfheal_built="${{ steps.selfheal.outputs.built }}"
|
||||
baseline_built="${{ steps.baseline_build.outputs.built }}"
|
||||
candidate_built="${{ steps.candidate_build.outputs.built }}"
|
||||
|
||||
args=(--duration "$duration" --rounds 3 --cooldown 5 --health-timeout 180 --baseline-revision "$baseline_sha" --candidate-revision "$candidate_sha")
|
||||
|
||||
if [[ "$baseline_hit" == "true" || "$selfheal_built" == "true" || "$baseline_built" == "true" ]]; then
|
||||
chmod +x baseline-bin/rustfs
|
||||
base_bin="$PWD/baseline-bin/rustfs"
|
||||
args+=(--baseline-bin "$base_bin")
|
||||
if [[ "$baseline_hit" == "true" ]]; then
|
||||
base_src="actions-cache (rustfs-baseline-$baseline_sha)"
|
||||
elif [[ "$selfheal_built" == "true" ]]; then
|
||||
base_src="source build (cache self-heal, saved as rustfs-baseline-$baseline_sha)"
|
||||
else
|
||||
base_src="isolated baseline source build (saved as rustfs-baseline-$baseline_sha)"
|
||||
fi
|
||||
if [[ "$candidate_sha" == "$baseline_sha" ]]; then
|
||||
# No commits landed since the last successful baseline, so reuse
|
||||
# the one binary for both phases and measure only rig drift.
|
||||
args+=(--candidate-bin "$base_bin")
|
||||
cand_src="same binary as baseline (same commit)"
|
||||
elif [[ "$candidate_built" == "true" ]]; then
|
||||
chmod +x candidate-bin/rustfs
|
||||
args+=(--candidate-bin "$PWD/candidate-bin/rustfs")
|
||||
cand_src="source build of the checked-out ref"
|
||||
else
|
||||
echo "::error::candidate binary was not built" >&2
|
||||
exit 2
|
||||
fi
|
||||
else
|
||||
echo "::error::baseline binary was not restored or built" >&2
|
||||
exit 2
|
||||
fi
|
||||
|
||||
echo "baseline binary: $base_src"
|
||||
echo "candidate binary: $cand_src"
|
||||
|
||||
if [[ "${{ steps.exempt.outputs.allow_regression }}" == "true" ]]; then
|
||||
args+=(--allow-regression --exemption-reason "workflow dispatch override")
|
||||
fi
|
||||
# Do not let a gate FAIL abort the job here; capture status and surface
|
||||
# it after the step summary is written.
|
||||
set +e
|
||||
bash scripts/run_hotpath_warp_abba.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-abba/*/ 2>/dev/null | head -n1 || true)"
|
||||
echo "run_dir=${run_dir%/}" >> "$GITHUB_OUTPUT"
|
||||
# shellcheck disable=SC2012
|
||||
gate_md="$(ls -t target/hotpath-abba/*/candidate_gate.md 2>/dev/null | head -n1 || true)"
|
||||
echo "gate_md=$gate_md" >> "$GITHUB_OUTPUT"
|
||||
|
||||
- name: Upload A/B results
|
||||
if: always()
|
||||
uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
|
||||
with:
|
||||
name: hotpath-warp-ab-${{ github.run_number }}
|
||||
# Includes per-cell median_summary.csv / baseline_compare.csv, both gates,
|
||||
# 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-abba/
|
||||
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
|
||||
} >> "$GITHUB_STEP_SUMMARY"
|
||||
|
||||
- name: Stage successful candidate baseline
|
||||
if: >-
|
||||
steps.ab.outputs.status == '0' &&
|
||||
steps.commits.outputs.baseline_sha != steps.commits.outputs.candidate_sha
|
||||
run: |
|
||||
set -euo pipefail
|
||||
cp candidate-bin/rustfs baseline-bin/rustfs
|
||||
|
||||
- name: Cache successful candidate baseline
|
||||
if: >-
|
||||
steps.ab.outputs.status == '0' &&
|
||||
steps.commits.outputs.baseline_sha != steps.commits.outputs.candidate_sha
|
||||
uses: actions/cache/save@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6
|
||||
with:
|
||||
path: baseline-bin/rustfs
|
||||
key: rustfs-baseline-${{ steps.commits.outputs.candidate_sha }}
|
||||
|
||||
# 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="${{ steps.ab.outputs.status }}"
|
||||
if [[ -z "$status" ]]; then
|
||||
echo "::error::warp A/B setup failed before the rig ran. Check the first failed workflow step." >&2
|
||||
exit 1
|
||||
fi
|
||||
if [[ "$status" != "0" ]]; then
|
||||
echo "::error::warp A/B budget gate failed (exit $status). See the step summary / gate.md artifact." >&2
|
||||
exit "$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); 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.
|
||||
if: >-
|
||||
github.repository == 'rustfs/rustfs' && (
|
||||
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@f548e57e544e1ff5a4c46bf1e1b8685f8e4a348a # v7
|
||||
with:
|
||||
persist-credentials: false
|
||||
- name: Open or update failure-tracking issue
|
||||
uses: ./.github/actions/schedule-failure-issue
|
||||
with:
|
||||
github-token: ${{ secrets.BACKLOG_ISSUE_TOKEN }}
|
||||
@@ -16,8 +16,8 @@
|
||||
#
|
||||
# Forces a job failure and then runs .github/actions/schedule-failure-issue
|
||||
# through the exact `needs` + `always() && contains(needs.*.result,
|
||||
# 'failure')` wiring used by the real consumers (e2e-s3tests, mint, fuzz,
|
||||
# performance-ab). Dispatch it twice to verify both the issue-creation and
|
||||
# 'failure')` wiring used by the real consumers (e2e-s3tests, mint, fuzz).
|
||||
# Dispatch it twice to verify both the issue-creation and
|
||||
# the dedupe-comment paths, then close the resulting
|
||||
# "[scheduled-failure] Schedule Failure Alert Drill" issue in rustfs/backlog
|
||||
# (the default target repository). Requires the BACKLOG_ISSUE_TOKEN secret
|
||||
|
||||
@@ -28,7 +28,6 @@ on:
|
||||
- "mint"
|
||||
- "minio-interop"
|
||||
- "Nightly GNU Build"
|
||||
- "Performance A/B"
|
||||
- "Runner Hygiene"
|
||||
types: [completed]
|
||||
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
# Hotpath warp runbook (A/B gate and ABBA evidence)
|
||||
|
||||
**Use this when:** you need S3-face (PutObject / GetObject / mixed) performance evidence for a code change: the nightly relative-budget gate, a quick local A/B, or a formal ABBA run whose numbers will be quoted in a PR.
|
||||
**Source of truth:** `scripts/run_hotpath_warp_ab.sh` (quick A/B rig), `scripts/run_hotpath_warp_abba.sh` (ABBA runner; also what CI executes), `scripts/hotpath_warp_ab_gate.sh` (relative-budget gate), `scripts/run_object_batch_bench_enhanced.sh` (warp driver, medians, `baseline_compare.csv`), `.github/workflows/performance-ab.yml` (CI gate).
|
||||
**Use this when:** you need S3-face (PutObject / GetObject / mixed) performance evidence for a code change: a quick local A/B, or a formal ABBA run whose numbers will be quoted in a PR.
|
||||
**Source of truth:** `scripts/run_hotpath_warp_ab.sh` (quick A/B rig), `scripts/run_hotpath_warp_abba.sh` (ABBA runner), `scripts/hotpath_warp_ab_gate.sh` (relative-budget gate), `scripts/run_object_batch_bench_enhanced.sh` (warp driver, medians, `baseline_compare.csv`).
|
||||
|
||||
This runbook compares two binaries. To sweep one `RUSTFS_*` runtime knob at a time against a fixed binary, use [object-io-tuning-ab-matrix.md](object-io-tuning-ab-matrix.md) instead.
|
||||
|
||||
@@ -10,7 +10,7 @@ Two entry points share one workload matrix, one gate, and one deploy-hook shape:
|
||||
| Entry | Script | Legs per cell | Use |
|
||||
| --- | --- | --- | --- |
|
||||
| Quick A/B | `scripts/run_hotpath_warp_ab.sh` | baseline → candidate | local smoke, fast triage of a suspected regression |
|
||||
| ABBA | `scripts/run_hotpath_warp_abba.sh` | A1 baseline → B1 candidate → B2 candidate → A2 baseline | formal evidence; the CI gate |
|
||||
| ABBA | `scripts/run_hotpath_warp_abba.sh` | A1 baseline → B1 candidate → B2 candidate → A2 baseline | formal evidence |
|
||||
|
||||
This rig is the only entry point for S3-face performance coverage. There is deliberately no in-process criterion benchmark for those operations: a criterion harness that stands up an embedded server measures the harness, not the S3 path. Micro-benchmarks stay at function level (EC encode, `xl.meta` parse, `rename_data`). Knob-level sweeps against a fixed binary are a different question; see [object-io-tuning-ab-matrix.md](object-io-tuning-ab-matrix.md).
|
||||
|
||||
@@ -163,16 +163,6 @@ After each window confirm the `.json.gz` profile and its `.syms.json` sidecar ar
|
||||
|
||||
Instrumented builds (features in `rustfs/Cargo.toml`): `--features hotpath-alloc` for allocation attribution, `--features hotpath-cpu` for CPU hotpath sections. Compare instrumented binaries only with other builds of the same mode; never use them for throughput acceptance, because the instrumentation changes what is measured.
|
||||
|
||||
## CI gate (`performance-ab.yml`)
|
||||
## Performance gate status
|
||||
|
||||
| Aspect | Value |
|
||||
| --- | --- |
|
||||
| Triggers | `schedule` (nightly cron `31 6 * * *` UTC against `main`) and `workflow_dispatch` (inputs `duration`, default `12s`; `allow_regression`, boolean). No `pull_request` trigger and no label gating. |
|
||||
| Jobs | `warp-ab` (runner `sm-standard-2`, `timeout-minutes: 180`); `alert-on-failure` (opens the scheduled-failure issue; scheduled runs only). |
|
||||
| Baseline commit | scheduled: head of the last successful scheduled run (falls back to the candidate itself when there is none); dispatch: `origin/main`. The baseline must be an ancestor of the candidate. |
|
||||
| Binary cache | `actions/cache` inside `warp-ab`, key `rustfs-baseline-<baseline_sha>`. Miss: source build (same-commit runs build once and reuse the binary for both phases); a run whose gate passes saves the candidate binary under `rustfs-baseline-<candidate_sha>` for the next night. |
|
||||
| Command | `scripts/run_hotpath_warp_abba.sh --duration <input> --rounds 3 --cooldown 5 --health-timeout 180 --baseline-revision <sha> --candidate-revision <sha> --baseline-bin ... --candidate-bin ...` |
|
||||
| Exemption | `allow_regression=true` on dispatch adds `--allow-regression --exemption-reason "workflow dispatch override"`. |
|
||||
| Artifacts | `hotpath-warp-ab-<run_number>` containing `target/hotpath-abba/` (14-day retention); the step summary renders `candidate_gate.md`, or the server-log tails when the rig failed before the gate. The `Enforce gate` step fails the job on a non-zero rig exit. |
|
||||
|
||||
The nightly detects a regression within a day of landing; it does not block a merge. For pre-merge evidence run the ABBA procedure above and attach the outputs to the PR.
|
||||
There is no scheduled CI performance gate. The ABBA procedure above is the evidence path for performance comparisons; attach its outputs to the PR that claims a performance effect. A scheduled regression gate requires calibrated budgets: on the previous gate's runner, measured drift between identical binaries at the default 12-second rounds exceeded the former 10% fail budget, so the scheduled entry never seeded a baseline (the workflow only ever passed on its pre-removal push trigger). Restore a scheduled gate only after a noise study grounds the budgets on the target runner.
|
||||
|
||||
@@ -41,7 +41,7 @@ The `--abba` mode runs five independent scenario cells: `cold-hot`, `fresh-hot`,
|
||||
three A1/B1/B2/A2 groups for both baseline/candidate with background work on,
|
||||
and candidate-only background off/on. A measured release leg lasts at least 7200
|
||||
seconds; the minimum matrix contains 120 legs (240 hours before setup/oracles).
|
||||
The existing `performance-ab.yml` supplies the pattern for immutable build
|
||||
The hotpath warp ABBA rig (`scripts/run_hotpath_warp_abba.sh`) supplies the pattern for immutable build
|
||||
provenance and failure propagation, but its short Warp workload is not this
|
||||
scanner gate. No scheduled workflow starts this matrix automatically.
|
||||
|
||||
|
||||
@@ -99,7 +99,6 @@ Scheduled lanes never block a PR. Their workflow-local gate fails the run, sched
|
||||
| `fuzz.yml` (nightly) | `nightly-fuzz-corpus` per target | gate; corpus and crash artifacts | yes | `MAX_TOTAL_TIME=<seconds> ./scripts/fuzz/run.sh` |
|
||||
| `minio-interop.yml` (nightly) | `minio-interop` | EC + SSE read-parity gate | yes, with `never_ran_grace_until` | pinned Docker fixture steps in the workflow |
|
||||
| `on-demand-migration-interop.yml` (nightly) | `minio-source`, `cloud-source` (`aws`, `r2`, `gcs`) | report-only provider interop; one JSON report per provider naming cases, timings and source request counts, plus JUnit and MinIO logs. A cloud provider whose `ODM_INTEROP_*` secrets are absent is skipped with a summary note, not failed | no | start the pinned MinIO container as in the job, export the `RUSTFS_ODM_INTEROP_*` variables, run `python3 scripts/e2e_binary.py build --bins`, then `python3 scripts/e2e_binary.py run -- cargo nextest run --profile e2e-odm-interop -p e2e_test` |
|
||||
| `performance-ab.yml` (nightly) | `warp-ab` | regression-budget gate; A/B summaries and server logs | yes | `bash scripts/run_hotpath_warp_abba.sh --help` |
|
||||
| `nightly-gnu.yml` (nightly) | `build`, `kms-vault-lane`, `kms-vault-ha-failover` | build, live Vault, and HA failover gates | yes | commands and pinned Vault images in the workflow |
|
||||
| `audit.yml` (nightly) | `cargo-deny`, `workflow-pin-report` | dependency and workflow-pin gates | yes | `cargo deny check`; `scripts/security/check_workflow_pins.sh` |
|
||||
| `mint.yml` (weekly) | `mint` | report-only by design; per-suite PASS/FAIL/NA and raw `log.json` | yes | pinned Docker sequence in the workflow |
|
||||
@@ -111,7 +110,7 @@ Scheduled lanes never block a PR. Their workflow-local gate fails the run, sched
|
||||
| `gateway-integration-sync.yml` (weekdays) | `sync` | merges `main` into `gateway/integration` as a `--no-ff` merge commit; a conflicting merge is aborted and opens a `sync/<source>-<date>` pull request listing the conflicting files, which keeps the run green; a rejected push or missing pull request fails it | no | dispatch, optionally with `source_ref` |
|
||||
| `scheduled-validation-freshness.yml` (nightly) | `check-freshness` | fails on missing or stale attempts or completed successes | n/a | dispatch |
|
||||
|
||||
Manual `workflow_dispatch` runs are debugging evidence and do not open scheduled-failure issues. A manual performance run may explicitly allow a known regression; that override is not a passing baseline.
|
||||
Manual `workflow_dispatch` runs are debugging evidence and do not open scheduled-failure issues.
|
||||
|
||||
## Packaged functional acceptance
|
||||
|
||||
|
||||
+1
-1
@@ -100,7 +100,7 @@ their issue closes.
|
||||
| Entry | Status | Purpose | Wiring / docs |
|
||||
|---|---|---|---|
|
||||
| `run_hotpath_warp_ab.sh` | dev-tool | Linux warp A/B rig for the hotpath series (quick local A/B) | `docs/operations/hotpath-warp-ab-runbook.md` |
|
||||
| `run_hotpath_warp_abba.sh` | ci-gate | Formal ABBA warp runner (baseline/candidate interleaved, `--rounds >= 3`); the runner performance-ab.yml executes | performance-ab.yml (scheduled); `docs/operations/hotpath-warp-ab-runbook.md` |
|
||||
| `run_hotpath_warp_abba.sh` | dev-tool | Formal ABBA warp runner (baseline/candidate interleaved, `--rounds >= 3`) | `docs/operations/hotpath-warp-ab-runbook.md` |
|
||||
| `hotpath_warp_ab_gate.sh` | dev-tool | Relative-budget gate evaluated over the warp A/B results | used by `run_hotpath_warp_ab.sh`; hotpath runbook |
|
||||
| `run_internode_grpc_ab_bench.sh` | dev-tool | One-click A/B driver for the internode gRPC optimization stages | `docs/operations/internode-grpc-benchmark-runbook.md` |
|
||||
| `run_internode_transport_baseline.sh` | dev-tool | Internode transport baseline runner | internode runbook; `crates/io-metrics/README.md` |
|
||||
|
||||
@@ -520,7 +520,6 @@ SCHEDULED_ALERT_CHECKOUT_OVERRIDES = {
|
||||
".github/workflows/ci.yml": "actions/checkout@f548e57e544e1ff5a4c46bf1e1b8685f8e4a348a",
|
||||
".github/workflows/fuzz.yml": "actions/checkout@f548e57e544e1ff5a4c46bf1e1b8685f8e4a348a",
|
||||
".github/workflows/nightly-gnu.yml": "actions/checkout@f548e57e544e1ff5a4c46bf1e1b8685f8e4a348a",
|
||||
".github/workflows/performance-ab.yml": "actions/checkout@f548e57e544e1ff5a4c46bf1e1b8685f8e4a348a",
|
||||
".github/workflows/runner-hygiene.yml": "actions/checkout@f548e57e544e1ff5a4c46bf1e1b8685f8e4a348a",
|
||||
".github/workflows/scheduled-validation-watchdog.yml": "actions/checkout@f548e57e544e1ff5a4c46bf1e1b8685f8e4a348a",
|
||||
".github/workflows/scheduled-validation-freshness.yml": "actions/checkout@f548e57e544e1ff5a4c46bf1e1b8685f8e4a348a",
|
||||
@@ -5953,7 +5952,6 @@ class SelfTests(unittest.TestCase):
|
||||
".github/workflows/ci.yml",
|
||||
".github/workflows/fuzz.yml",
|
||||
".github/workflows/nightly-gnu.yml",
|
||||
".github/workflows/performance-ab.yml",
|
||||
".github/workflows/runner-hygiene.yml",
|
||||
)
|
||||
with tempfile.TemporaryDirectory() as tmp:
|
||||
|
||||
@@ -1,43 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
set -euo pipefail
|
||||
|
||||
workflow=".github/workflows/performance-ab.yml"
|
||||
|
||||
require_absent_pattern() {
|
||||
local pattern="$1"
|
||||
local description="$2"
|
||||
|
||||
if grep -Eq -- "$pattern" "$workflow"; then
|
||||
echo "invalid performance A/B workflow contract: $description" >&2
|
||||
exit 1
|
||||
fi
|
||||
}
|
||||
|
||||
require_present_pattern() {
|
||||
local pattern="$1"
|
||||
local description="$2"
|
||||
|
||||
if ! grep -Eq -- "$pattern" "$workflow"; then
|
||||
echo "invalid performance A/B workflow contract: $description" >&2
|
||||
exit 1
|
||||
fi
|
||||
}
|
||||
|
||||
require_absent_pattern '(^|[^[:alnum:]_])pull_request(_target)?([^[:alnum:]_]|$)' "the workflow must not contain PR event handling"
|
||||
require_absent_pattern 'pull-requests[[:space:]]*:[[:space:]]*write' "the workflow must not receive PR write permission"
|
||||
require_absent_pattern 'permissions[[:space:]]*:[[:space:]]*write-all' "the workflow must not receive broad write permission"
|
||||
require_absent_pattern '^[[:space:]]*push:' "the workflow must not spend a release build on every main push"
|
||||
require_present_pattern 'listWorkflowRuns' "the scheduled baseline must come from workflow history"
|
||||
require_present_pattern 'status:[[:space:]]*"success"' "the scheduled baseline must be a successful run"
|
||||
require_present_pattern 'SCHEDULED_BASELINE_SHA' "the resolved scheduled baseline must reach the comparison"
|
||||
require_present_pattern "SCHEDULED_BASELINE_SHA:-\\\$candidate_sha" "the first scheduled run must seed from its verified candidate"
|
||||
require_present_pattern 'git merge-base --is-ancestor' "every baseline must stay on candidate history"
|
||||
require_present_pattern 'update the selected ref before comparing' "stale manual refs must fail before the benchmark"
|
||||
require_present_pattern 'setup failed before the rig ran' "an earlier setup failure must not become an empty exit status"
|
||||
require_present_pattern 'Cache successful candidate baseline' "a successful candidate must become the next cached baseline"
|
||||
if ! sed -n '/^ warp-ab:/,/^ alert-on-failure:/p' "$workflow" | grep -Eq '^ timeout-minutes:[[:space:]]*180([[:space:]]|$)'; then
|
||||
echo "invalid performance A/B workflow contract: the cold-cache path must fit both builds, the A/B run, and evidence publication" >&2
|
||||
exit 1
|
||||
fi
|
||||
|
||||
echo "Performance A/B workflow contract ok."
|
||||
Reference in New Issue
Block a user