Every path-gated workflow begins with one small job -- `does this change touch
it`, or `which workspaces this change touches` for the standards build -- that
all of its real jobs declare in `needs`. All 55 of those gate jobs ran on
`linux-arm64`.
That is the one label the fleet cannot spare. Sampled while 267 runs were
queued, every waiting run had the same shape:
[waiting] linux-arm64 does this change touch it
[blocked] linux member · dataflow device, simulated
[blocked] macos member · dataflow device absent
237 runs, each holding a gate in the longest queue, and every real job blocked
behind it. Not only `linux` jobs: `macos`, `linux-hw`, `linux-arm64-hw` and
`macos-m3-ultra` jobs were blocked too. The measured effect was `linux-arm64`
263 deep against `linux` **zero**, with jetson-hub's 20 cores taking no job for
a full hour -- starved not for want of work but because the work could not be
unblocked. A gate waiting on the resource it is rationing is head-of-line
blocking, and it makes every other runner in the fleet idle on its behalf.
The gates move to `linux`. Nothing else changes: the diff is 55 `runs-on` lines
and no others, and every touched file still parses.
Why `linux` and not elsewhere. The gate is not free -- it compares
`BASE...HEAD`, which needs the merge base, so `fetch-depth: 0` is a full clone
rather than a ten-second script. `macos` is served by hosts on a Mac Studio
sitting at load 185; `linux` is a separate 20-core box that was measurably
empty. This moves a serialisation point to the least contended label; it does
not remove it, and the gate is still a single point of failure for its
workflow. The structural fix is an early-exit first step inside each job, which
costs one checkout per leg and is a larger change.
Follow-up, deliberately not bundled: jetson-hub runs at capacity 2 with 15 GB of
RAM, so 55 clones still serialise two at a time. Raising capacity helps gates
and hurts the heavy emulated-architecture jobs that share the label, so it wants
measuring on its own.
Co-Authored-By: Claude Opus 5 <[email protected]>
162 lines
7.2 KiB
YAML
162 lines
7.2 KiB
YAML
name: wai-quantum views in agent hosts
|
|
|
|
# The engine's answers drawn as interactive views inside an agent host, through
|
|
# the Model Context Protocol's extension for them (wai/wai-quantum/src/engine/views.rs,
|
|
# the page in wai/wai-quantum/views/, the engine inside it in wai/wai-quantum-views/,
|
|
# the views themselves in wai/quantum-studio/src/views/):
|
|
# - the committed page is built from its sources as they are (build-views.mjs
|
|
# --check), and rebuilt here: its script and styles to the same bytes. The
|
|
# engine inside it is compiled from wai-quantum, and LLVM lays the same code
|
|
# out differently once anything else in that crate changes, so its bytes are
|
|
# compared and a difference is a warning to rebuild, not a failure: the
|
|
# scripted host below holds the committed page's engine to the receipts the
|
|
# current servers seal;
|
|
# - the engine inside the views checks a receipt as the engine does, and
|
|
# engine::views gives the resources, the tool map, each view's data and its
|
|
# answer in plain words, every view within the size limit;
|
|
# - a scripted host draws every view from both MCP servers in a headless
|
|
# browser (test/views.test.mjs): under the extension's default
|
|
# content-security policy, which refuses WebAssembly, the receipt is checked
|
|
# in the view by the engine translated to JavaScript, and with WebAssembly
|
|
# allowed, by the WebAssembly; a changed result fails the check; the view's
|
|
# own tool for a running job is kept from the model; no view makes a request.
|
|
|
|
concurrency:
|
|
group: ${{ github.workflow }}-${{ github.ref }}
|
|
cancel-in-progress: true
|
|
|
|
on:
|
|
push:
|
|
branches: [main]
|
|
paths:
|
|
- 'wai/wai-quantum-views/**'
|
|
- 'wai/wai-quantum/**'
|
|
- 'wai/wai-quantum-cli/**'
|
|
- 'wai/wai-quantum-service/**'
|
|
- 'wai/wai-quantum-qpu/**'
|
|
- 'wai/quantum-studio/**'
|
|
- 'rust-toolchain.toml'
|
|
- '.github/workflows/wai-quantum-views.yml'
|
|
pull_request:
|
|
paths:
|
|
- 'wai/wai-quantum-views/**'
|
|
- 'wai/wai-quantum/**'
|
|
- 'wai/wai-quantum-cli/**'
|
|
- 'wai/wai-quantum-service/**'
|
|
- 'wai/wai-quantum-qpu/**'
|
|
- 'wai/quantum-studio/**'
|
|
- 'rust-toolchain.toml'
|
|
- '.github/workflows/wai-quantum-views.yml'
|
|
|
|
jobs:
|
|
relevant:
|
|
name: does this change touch it
|
|
runs-on: linux
|
|
outputs:
|
|
run: ${{ steps.gate.outputs.run }}
|
|
steps:
|
|
- uses: actions/checkout@v5
|
|
with:
|
|
fetch-depth: 0
|
|
- id: gate
|
|
run: |
|
|
set -euo pipefail
|
|
r=$(python3 scripts/pr-touches-paths.py .github/workflows/wai-quantum-views.yml \
|
|
"${{ github.event.pull_request.base.sha }}" \
|
|
"${{ github.event.pull_request.head.sha }}")
|
|
echo "this change touches the views: $r"
|
|
echo "run=$r" >> "$GITHUB_OUTPUT"
|
|
|
|
page:
|
|
name: views · the page and its engine
|
|
needs: relevant
|
|
if: needs.relevant.outputs.run == 'true'
|
|
runs-on: linux
|
|
timeout-minutes: 60
|
|
steps:
|
|
- uses: actions/checkout@v5
|
|
- name: Install Rust 1.98.0 with the wasm32 target
|
|
shell: bash
|
|
run: |
|
|
export PATH="$HOME/.cargo/bin:$PATH"
|
|
command -v rustup >/dev/null 2>&1 || \
|
|
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --profile minimal --default-toolchain 1.98.0
|
|
rustup toolchain install 1.98.0 --profile minimal
|
|
rustup default 1.98.0
|
|
rustup target add wasm32-unknown-unknown
|
|
echo "$HOME/.cargo/bin" >> "$GITHUB_PATH"
|
|
- uses: actions/setup-node@v4
|
|
with:
|
|
node-version: '22'
|
|
- name: The committed page is built from these sources
|
|
shell: bash
|
|
working-directory: wai/quantum-studio
|
|
run: |
|
|
set -euo pipefail
|
|
npm ci --no-audit --no-fund
|
|
node build-views.mjs --check
|
|
- name: The engine inside the views, and engine::views
|
|
shell: bash
|
|
run: |
|
|
set -euo pipefail
|
|
(cd wai/wai-quantum-views && cargo test --locked --release 2>&1 | tee "$RUNNER_TEMP/views.log")
|
|
grep -qE '^test result: ok\. 3 passed; 0 failed' "$RUNNER_TEMP/views.log" || { echo "::error::the views' engine tests did not all run and pass"; exit 1; }
|
|
(cd wai/wai-quantum && cargo test --locked --release --features full --lib engine::views:: 2>&1 | tee "$RUNNER_TEMP/engine.log")
|
|
grep -qE '^test result: ok\. 4 passed; 0 failed' "$RUNNER_TEMP/engine.log" || { echo "::error::engine::views' tests did not all run and pass"; exit 1; }
|
|
- name: Rebuilt here, its script and styles to the same bytes
|
|
shell: bash
|
|
run: |
|
|
set -euo pipefail
|
|
bash wai/wai-quantum-views/build.sh "$RUNNER_TEMP/views"
|
|
python3 - "$RUNNER_TEMP/views/manifest.json" wai/wai-quantum/views/manifest.json <<'EOF'
|
|
import json, sys
|
|
new, old = (json.load(open(p)) for p in sys.argv[1:])
|
|
bad = [k for k in ("script", "styles") if new["sha256"][k] != old["sha256"][k]]
|
|
if bad:
|
|
print(f"::error::the committed views page's {', '.join(bad)} are not what their sources build: run bash wai/wai-quantum-views/build.sh and commit what it writes")
|
|
sys.exit(1)
|
|
for k in ("engine_wasm", "engine_js"):
|
|
if new["sha256"][k] != old["sha256"][k]:
|
|
print(f"::warning::the engine inside the views ({k}) compiles to other bytes than the committed page holds: wai-quantum changed since it was built; rebuild it with bash wai/wai-quantum-views/build.sh")
|
|
print("the page's script and styles are what their sources build")
|
|
EOF
|
|
|
|
hosts:
|
|
name: views · a scripted host
|
|
needs: relevant
|
|
if: needs.relevant.outputs.run == 'true'
|
|
runs-on: linux
|
|
timeout-minutes: 75
|
|
steps:
|
|
- uses: actions/checkout@v5
|
|
- name: Install Rust 1.98.0
|
|
shell: bash
|
|
run: |
|
|
export PATH="$HOME/.cargo/bin:$PATH"
|
|
command -v rustup >/dev/null 2>&1 || \
|
|
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --profile minimal --default-toolchain 1.98.0
|
|
rustup toolchain install 1.98.0 --profile minimal
|
|
rustup default 1.98.0
|
|
echo "$HOME/.cargo/bin" >> "$GITHUB_PATH"
|
|
- uses: actions/setup-node@v4
|
|
with:
|
|
node-version: '22'
|
|
- name: Both servers
|
|
shell: bash
|
|
run: |
|
|
set -euo pipefail
|
|
(cd wai/wai-quantum-cli && cargo build --locked --release)
|
|
(cd wai/wai-quantum-service && cargo build --locked --release)
|
|
- name: Every view drawn by a scripted host in a browser
|
|
shell: bash
|
|
working-directory: wai/quantum-studio
|
|
env:
|
|
WAI_E2E_CHANNEL: ''
|
|
run: |
|
|
set -euo pipefail
|
|
npm ci --no-audit --no-fund
|
|
npx playwright install --with-deps chromium
|
|
WQ_CLI="$GITHUB_WORKSPACE/wai/wai-quantum-cli/target/release/wai-quantum" \
|
|
WQ_SERVICE="$GITHUB_WORKSPACE/wai/wai-quantum-service/target/release/wai-quantum-service" \
|
|
node test/views.test.mjs 2>&1 | tee "$RUNNER_TEMP/views.log"
|
|
grep -q '^every check passed$' "$RUNNER_TEMP/views.log" || { echo "::error::the scripted host's checks did not all pass"; exit 1; }
|