ci: try using gha

Signed-off-by: Tiago Castro <[email protected]>

.github/workflows/bdd.yml

@@ -0,0 +1,53 @@
+name: BDD CI
+on:
+  workflow_call:
+
+env:
+  CARGO_TERM_COLOR: always
+  CARGO_INCREMENTAL: 0
+
+jobs:
+  bdd-tests:
+    runs-on: ubuntu-latest-16-cores
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          submodules: 'recursive'
+      - uses: DeterminateSystems/nix-installer-action@v14
+      - uses: DeterminateSystems/magic-nix-cache-action@v8
+      - name: Pre-populate nix-shell
+        run: |
+          export NIX_PATH=nixpkgs=$(jq '.nixpkgs.url' nix/sources.json -r)
+          echo "NIX_PATH=$NIX_PATH" >> $GITHUB_ENV
+          nix-shell --run "echo" shell.nix
+      - name: Handle Rust dependencies caching
+        uses: Swatinem/rust-cache@v2
+      - name: Build binaries
+        run: nix-shell --run "cargo build --bins"
+      - name: Setup Test Pre-Requisites
+        run: |
+          sudo sysctl -w vm.nr_hugepages=3072
+          sudo apt-get install linux-modules-extra-$(uname -r)
+          sudo modprobe nvme_tcp
+      - name: Run BDD Tests
+        run: |
+          nix-shell --run "deployer start --image-pull-policy always -w 60s && deployer stop"
+          nix-shell --run "./scripts/python/test.sh"
+      - name: Cleanup
+        if: always()
+        run: nix-shell --run "./scripts/python/test-residue-cleanup.sh"
+      - name: Surface failing tests
+        if: always()
+        uses: pmeier/pytest-results-action@main
+        with:
+          path: report.xml
+          summary: true
+          display-options: a
+          fail-on-empty: true
+          title: Test results
+# debugging
+      - name: Setup tmate session
+        if: ${{ failure() }}
+        timeout-minutes: 120
+        uses: mxschmitt/action-tmate@v3

.github/workflows/image-pr.yml

@@ -0,0 +1,19 @@
+name: ImageBuild CI
+on:
+  workflow_call:
+
+env:
+  CARGO_TERM_COLOR: always
+  CARGO_INCREMENTAL: 0
+
+jobs:
+  image-build-test:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: 'recursive'
+      - uses: DeterminateSystems/nix-installer-action@v14
+      - uses: DeterminateSystems/magic-nix-cache-action@v8
+      - name: Test building the release images
+        run: ./scripts/release.sh --skip-publish --build-bins

.github/workflows/image.yml

@@ -0,0 +1,24 @@
+name: Image Push
+on:
+  push:
+    branches:
+      - develop
+      - 'release/**'
+    tags:
+      - 'v[0-9]+.[0-9]+.[0-9]+**'
+
+env:
+  CARGO_TERM_COLOR: always
+  CARGO_INCREMENTAL: 0
+
+jobs:
+  image-build-test:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: 'recursive'
+      - uses: DeterminateSystems/nix-installer-action@v14
+      - uses: DeterminateSystems/magic-nix-cache-action@v8
+      - name: Test building the release images
+        run: ./scripts/release.sh

.github/workflows/lint.yml

@@ -0,0 +1,26 @@
+name: Unit/Int CI
+on:
+  workflow_call:
+
+env:
+  CARGO_TERM_COLOR: always
+  CARGO_INCREMENTAL: 0
+
+jobs:
+  rust-linter:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: 'recursive'
+      - uses: DeterminateSystems/nix-installer-action@v14
+      - uses: DeterminateSystems/magic-nix-cache-action@v8
+      - name: Pre-populate nix-shell
+        run: |
+          export NIX_PATH=nixpkgs=$(jq '.nixpkgs.url' nix/sources.json -r)
+          echo "NIX_PATH=$NIX_PATH" >> $GITHUB_ENV
+          nix-shell --run "echo" shell.nix
+      - name: Handle Rust dependencies caching
+        uses: Swatinem/rust-cache@v2
+      - name: Lint rust code
+        run: nix-shell --run "./scripts/rust/linter.sh"

.github/workflows/nightly-ci.yml

@@ -0,0 +1,24 @@
+name: Nightly CI
+on:
+  workflow_dispatch:
+
+jobs:
+  lint-ci:
+    uses: ./.github/workflows/lint.yml
+  int-ci:
+    uses: ./.github/workflows/unit-int.yml
+  bdd-ci:
+    uses: ./.github/workflows/bdd.yml
+  image-ci:
+    uses: ./.github/workflows/image-pr.yml
+  bors-ci:
+    if: ${{ success() }}
+    needs:
+      - lint-ci
+      - int-ci
+      - bdd-ci
+      - image-ci
+    runs-on: ubuntu-latest
+    steps:
+      - name: CI succeeded
+        run: exit 0

.github/workflows/pr-ci.yml

@@ -0,0 +1,28 @@
+name: Bors CI
+on:
+  push:
+    branches:
+      - staging
+      - trying
+      - ci
+
+jobs:
+  # lint-ci:
+  #   uses: ./.github/workflows/lint.yml
+  int-ci:
+    uses: ./.github/workflows/unit-int.yml
+  bdd-ci:
+    uses: ./.github/workflows/bdd.yml
+  # image-ci:
+  #   uses: ./.github/workflows/image-pr.yml
+  bors-ci:
+    if: ${{ success() }}
+    needs:
+      # - lint-ci
+      - int-ci
+      - bdd-ci
+      # - image-ci
+    runs-on: ubuntu-latest
+    steps:
+      - name: CI succeeded
+        run: exit 0

.github/workflows/unit-int.yml

@@ -0,0 +1,50 @@
+name: Integration CI
+on:
+  workflow_call:
+
+env:
+  CARGO_TERM_COLOR: always
+  CARGO_INCREMENTAL: 0
+
+jobs:
+  int-tests:
+    runs-on: ubuntu-latest-16-cores
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          submodules: 'recursive'
+      - uses: DeterminateSystems/nix-installer-action@v14
+      - uses: DeterminateSystems/magic-nix-cache-action@v8
+      - name: Pre-populate nix-shell
+        run: |
+          export NIX_PATH=nixpkgs=$(jq '.nixpkgs.url' nix/sources.json -r)
+          echo "NIX_PATH=$NIX_PATH" >> $GITHUB_ENV
+          nix-shell --run "echo" shell.nix
+      - name: Handle Rust dependencies caching
+        uses: Swatinem/rust-cache@v2
+        with:
+          save-if: ${{ startsWith(github.ref_name, 'release/') || github.ref_name == 'develop' || github.ref_name == 'ci' }}
+      - name: Build rust binaries
+        run: nix-shell --run "cargo build --bins"
+      - name: Build the tests
+        run: nix-shell --run "./scripts/rust/test.sh --no-run"
+      - name: Setup Test Pre-Requisites
+        run: |
+          sudo sysctl -w vm.nr_hugepages=2560
+          sudo apt-get install linux-modules-extra-$(uname -r)
+          sudo modprobe nvme_tcp
+      - name: Run Tests
+        run: |
+          # pre-pull the required container images
+          deployer start --image-pull-policy always -w 60s && deployer stop
+          # includes both unit and integration tests
+          nix-shell --run "./scripts/rust/test.sh"
+      - name: Cleanup
+        if: always()
+        run: nix-shell --run "./scripts/rust/deployer-cleanup.sh"
+# debugging
+      - name: Setup tmate session
+        if: ${{ failure() }}
+        timeout-minutes: 120
+        uses: mxschmitt/action-tmate@v3

control-plane/agents/src/bin/core/tests/controller/mod.rs

@@ -81,6 +81,7 @@ async fn store_lease_lock() {
         .await
         .unwrap();
 
+    tokio::time::sleep(std::time::Duration::from_secs(1)).await;
     let lease_ttl = std::time::Duration::from_secs(2);
     let _core_agent = Etcd::new_leased(
         ["0.0.0.0:2379"],

scripts/rust/test.sh

@@ -27,7 +27,7 @@ cleanup_handler() {
 }
 
 cleanup_handler >/dev/null
-trap cleanup_handler INT QUIT TERM HUP EXIT
+#trap cleanup_handler INT QUIT TERM HUP EXIT
 
 set -euxo pipefail
 

shell.nix

@@ -40,7 +40,6 @@ mkShell {
     pre-commit
     python3
     utillinux
-    rdma-core
     which
     paperclip
   ] ++ pkgs.lib.optional (!norust) rust
@@ -54,6 +53,7 @@ mkShell {
     tini
     udev
     lvm2
+    rdma-core
   ] ++ pkgs.lib.optional (system == "aarch64-darwin") darwin.apple_sdk.frameworks.Security;
 
   LIBCLANG_PATH = "${llvmPackages.libclang.lib}/lib";

tests/bdd/features/health_probes/test_readiness_probe.py

@@ -15,15 +15,18 @@
 
 def ready_http_get(context_msg_prefix: str):
     try:
-        response = requests.get("http://localhost:8081/ready", timeout=(0.003, 0.010))
+        response = requests.get("http://localhost:8081/ready", timeout=(0.1, 0.1))
         logger.info(
             f"{context_msg_prefix}: response.status_code: {response.status_code}"
         )
         return response
-    except requests.exceptions.Timeout:
+    except requests.exceptions.Timeout as e:
         logger.error(f"{context_msg_prefix}: the request timed out")
+        raise e
+
     except requests.exceptions.RequestException as e:
         logger.error(f"{context_msg_prefix}: an error occurred: {e}")
+        raise e
 
 
 @pytest.fixture(scope="module")