Merge branch 'develop' into cwirt/upstream-jni-scan-prefetch

Author: ChristopherWirt

.github/workflows/ci.yml

@@ -105,6 +105,7 @@ jobs:
             --with tabulate --with xxhash \
             pytest scripts/tests/test_measurement_id.py \
               scripts/tests/test_benchmark_reporting.py scripts/tests/test_post_ingest.py \
+              scripts/tests/test_refresh_bench_site.py \
               scripts/tests/test_codspeed_shards.py
 
   python-test:

.github/workflows/codspeed.yml

@@ -48,8 +48,8 @@ jobs:
               - ".github/workflows/codspeed.yml"
 
   # CodSpeed rejects an upload with too many benchmarks, so the simulation benchmarks run in
-  # shards of at most 800. The planner reads only Cargo.toml files and the committed benchmark
-  # counts, so this job needs no Rust toolchain and takes seconds.
+  # shards of at most 800. The planner balances recorded target runtimes across the available
+  # slots, keeping allocation modes separate. It reads only manifests and committed estimates.
   plan-codspeed:
     name: "Plan Codspeed shards"
     if: github.event.pull_request.draft != true

.github/workflows/commit-metadata.yml

@@ -47,8 +47,6 @@ jobs:
           RDS_BENCH_INSTANCE_ENDPOINT: ${{ vars.RDS_BENCH_INSTANCE_ENDPOINT }}
           RDS_BENCH_DB_NAME: ${{ vars.RDS_BENCH_DB_NAME }}
           AWS_REGION: ${{ vars.RDS_BENCH_REGION }}
-          BENCH_SITE_BASE_URL: ${{ vars.BENCH_SITE_BASE_URL }}
-          BENCH_REVALIDATE_TOKEN: ${{ secrets.BENCH_REVALIDATE_TOKEN }}
         run: |
           set -Eeuo pipefail
           echo -n > empty.jsonl
@@ -61,3 +59,13 @@ jobs:
             --commit-sha "${{ github.sha }}" \
             --branch "${{ github.ref_name }}" \
             --region "${AWS_REGION}"
+
+  # Tell the benchmarks website that the new commit exists. develop-bench.yml refreshes again
+  # after its benchmark results are ingested, but it does not wait for this workflow, so a push
+  # whose benchmark ingests all fail could otherwise leave this commit hidden by the cache.
+  refresh-site:
+    needs: commit-metadata
+    if: vars.GH_BENCH_INGEST_ROLE_ARN != ''
+    uses: ./.github/workflows/refresh-bench-site.yml
+    secrets:
+      BENCH_REVALIDATE_TOKEN: ${{ secrets.BENCH_REVALIDATE_TOKEN }}

.github/workflows/develop-bench.yml

@@ -14,11 +14,9 @@ permissions:
 jobs:
   build:
     timeout-minutes: 60
-    # GitHub drops runner labels longer than 255 characters, and the job then waits for a runner
-    # forever. With the long `family` list and the longest benchmark id, this label is close.
     runs-on: >-
       ${{ github.repository == 'vortex-data/vortex'
-          && format('runs-on={0}/region=us-east-1/runner=bench-dedicated/family=c8gd.8xlarge+c8g.8xlarge+c8gn.8xlarge+m8gd.8xlarge+m8g.8xlarge+r8gd.8xlarge+r8g.8xlarge+x8g.8xlarge+i8g.8xlarge/image=ubuntu26-full-arm64-pre-v2/extras=s3-cache/tag={1}', github.run_id, matrix.benchmark.id)
+          && format('runs-on={0}/region=us-east-1/runner=arm64-bench/extras=s3-cache/tag={1}', github.run_id, matrix.benchmark.id)
           || 'ubuntu-latest' }}
     strategy:
       fail-fast: false
@@ -238,8 +236,6 @@ jobs:
           RDS_BENCH_INSTANCE_ENDPOINT: ${{ vars.RDS_BENCH_INSTANCE_ENDPOINT }}
           RDS_BENCH_DB_NAME: ${{ vars.RDS_BENCH_DB_NAME }}
           AWS_REGION: ${{ vars.RDS_BENCH_REGION }}
-          BENCH_SITE_BASE_URL: ${{ vars.BENCH_SITE_BASE_URL }}
-          BENCH_REVALIDATE_TOKEN: ${{ secrets.BENCH_REVALIDATE_TOKEN }}
         run: |
           set -Eeuo pipefail
           curl -fsSL https://truststore.pki.rds.amazonaws.com/global/global-bundle.pem \
@@ -266,3 +262,13 @@ jobs:
     with:
       mode: "develop"
       matrix_preset: "develop"
+
+  # Tell the benchmarks website that new results exist, once per push, after every benchmark job
+  # of the push has ingested its results. `always()` runs it even when a benchmark job failed, so
+  # the results that other jobs ingested still appear.
+  refresh-site:
+    needs: [bench, sql]
+    if: ${{ always() }}
+    uses: ./.github/workflows/refresh-bench-site.yml
+    secrets:
+      BENCH_REVALIDATE_TOKEN: ${{ secrets.BENCH_REVALIDATE_TOKEN }}

.github/workflows/nightly-bench.yml

@@ -31,3 +31,12 @@ jobs:
         machine_type:
           - id: arm64
             instance_name: c8gd.metal-24xl
+
+  # Tell the benchmarks website that new results exist, once per run, after every machine type
+  # has ingested its results. `always()` runs it even when one machine type failed.
+  refresh-site:
+    needs: sql
+    if: ${{ always() }}
+    uses: ./.github/workflows/refresh-bench-site.yml
+    secrets:
+      BENCH_REVALIDATE_TOKEN: ${{ secrets.BENCH_REVALIDATE_TOKEN }}

.github/workflows/pr-bench-runner.yml

@@ -44,11 +44,9 @@ permissions:
 jobs:
   build:
     timeout-minutes: 60
-    # GitHub drops runner labels longer than 255 characters, and the job then waits for a runner
-    # forever. With the long `family` list and the longest benchmark id, this label is close.
     runs-on: >-
       ${{ github.repository == 'vortex-data/vortex'
-          && format('runs-on={0}/region=us-east-1/runner=bench-dedicated/family=c8gd.8xlarge+c8g.8xlarge+c8gn.8xlarge+m8gd.8xlarge+m8g.8xlarge+r8gd.8xlarge+r8g.8xlarge+x8g.8xlarge+i8g.8xlarge/image=ubuntu26-full-arm64-pre-v2/tag={1}{2}', github.run_id, inputs.benchmark_id, github.event.pull_request.head.repo.fork == false && '/extras=s3-cache' || '')
+          && format('runs-on={0}/region=us-east-1/runner=arm64-bench/tag={1}{2}', github.run_id, inputs.benchmark_id, github.event.pull_request.head.repo.fork == false && '/extras=s3-cache' || '')
           || 'ubuntu-latest' }}
     steps:
       - uses: runs-on/action@v2

.github/workflows/refresh-bench-site.yml

@@ -0,0 +1,29 @@
+# Tells the benchmarks website that new results exist, then warms its default pages. Callers run
+# it once per run, after their benchmark ingests finish, instead of once per benchmark job. This
+# keeps the website from flushing and refilling its cache dozens of times an hour while
+# benchmarks run. The script does nothing when the URL or the token is not set.
+
+name: Refresh benchmarks website
+
+on:
+  workflow_call:
+    secrets:
+      BENCH_REVALIDATE_TOKEN:
+        required: false
+
+jobs:
+  refresh:
+    if: vars.BENCH_SITE_BASE_URL != ''
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    permissions:
+      contents: read
+    steps:
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1  # v7
+        with:
+          sparse-checkout: scripts
+      - name: Revalidate and warm the benchmarks website
+        env:
+          BENCH_SITE_BASE_URL: ${{ vars.BENCH_SITE_BASE_URL }}
+          BENCH_REVALIDATE_TOKEN: ${{ secrets.BENCH_REVALIDATE_TOKEN }}
+        run: python3 scripts/refresh-bench-site.py

.github/workflows/sql-bench-matrix.yml

@@ -49,7 +49,7 @@ jobs:
       FLAT_LAYOUT_INLINE_ARRAY_NODE: "1"
     runs-on: >-
       ${{ github.repository == 'vortex-data/vortex'
-          && format('runs-on={0}/region=us-east-1/runner=bench-dedicated/family=c8gd.8xlarge+c8g.8xlarge+c8gn.8xlarge+m8gd.8xlarge+m8g.8xlarge+r8gd.8xlarge+r8g.8xlarge+x8g.8xlarge+i8g.8xlarge/image=ubuntu26-full-arm64-pre-v2/tag=build{1}', github.run_id, (inputs.mode != 'pr' || github.event.pull_request.head.repo.fork == false) && '/extras=s3-cache' || '')
+          && format('runs-on={0}/region=us-east-1/runner=arm64-bench/tag=build{1}', github.run_id, (inputs.mode != 'pr' || github.event.pull_request.head.repo.fork == false) && '/extras=s3-cache' || '')
           || 'ubuntu-latest' }}
     steps:
       - uses: runs-on/action@v2
@@ -308,8 +308,6 @@ jobs:
           RDS_BENCH_INSTANCE_ENDPOINT: ${{ vars.RDS_BENCH_INSTANCE_ENDPOINT }}
           RDS_BENCH_DB_NAME: ${{ vars.RDS_BENCH_DB_NAME }}
           AWS_REGION: ${{ vars.RDS_BENCH_REGION }}
-          BENCH_SITE_BASE_URL: ${{ vars.BENCH_SITE_BASE_URL }}
-          BENCH_REVALIDATE_TOKEN: ${{ secrets.BENCH_REVALIDATE_TOKEN }}
         run: |
           set -Eeuo pipefail
           curl -fsSL https://truststore.pki.rds.amazonaws.com/global/global-bundle.pem \

.github/workflows/web.yml

@@ -56,17 +56,19 @@ jobs:
         uses: taiki-e/install-action@1e29709c0ec63babbe8f266397a465f29a825f91  # v2
         with:
           tool: wasm-pack@0.15.0
+      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6  # v2.2.0
+        with:
+          bun-version: 1.4.3
+      # Storybook and Vite run on Node, and Bun runs their scripts with it.
       - uses: actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1  # v7
         with:
           node-version: "24"
-          cache: "npm"
-          cache-dependency-path: vortex-web/package-lock.json
-      - run: npm ci
-      - run: npm run wasm
-      - run: npm run format:check
-      - run: npm run lint
-      - run: npm run typecheck
-      - run: npm run build-storybook
+      - run: bun install --frozen-lockfile
+      - run: bun run wasm
+      - run: bun run format:check
+      - run: bun run lint
+      - run: bun run typecheck
+      - run: bun run build-storybook
 
   build:
     name: Build
@@ -83,13 +85,15 @@ jobs:
         uses: taiki-e/install-action@1e29709c0ec63babbe8f266397a465f29a825f91  # v2
         with:
           tool: wasm-pack@0.15.0
+      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6  # v2.2.0
+        with:
+          bun-version: 1.4.3
+      # Storybook and Vite run on Node, and Bun runs their scripts with it.
       - uses: actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1  # v7
         with:
           node-version: "24"
-          cache: "npm"
-          cache-dependency-path: vortex-web/package-lock.json
-      - run: npm ci
-      - run: npm run build
+      - run: bun install --frozen-lockfile
+      - run: bun run build
       - name: Upload build artifact
         uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9  # v7
         with: