|
| 1 | +# The nightly-gauntlet runs tests that are either too flaky or too slow to block |
| 2 | +# every PR. |
| 3 | +name: nightly-gauntlet |
| 4 | +on: |
| 5 | + schedule: |
| 6 | + # Every day at 4AM |
| 7 | + - cron: "0 4 * * 1-5" |
| 8 | + workflow_dispatch: |
| 9 | + |
| 10 | +permissions: |
| 11 | + contents: read |
| 12 | + |
| 13 | +jobs: |
| 14 | + test-go-pg: |
| 15 | + # make sure to adjust NUM_PARALLEL_PACKAGES and NUM_PARALLEL_TESTS below |
| 16 | + # when changing runner sizes |
| 17 | + runs-on: ${{ matrix.os == 'macos-latest' && github.repository_owner == 'coder' && 'depot-macos-latest' || matrix.os == 'windows-2022' && github.repository_owner == 'coder' && 'depot-windows-2022-16' || matrix.os }} |
| 18 | + # This timeout must be greater than the timeout set by `go test` in |
| 19 | + # `make test-postgres` to ensure we receive a trace of running |
| 20 | + # goroutines. Setting this to the timeout +5m should work quite well |
| 21 | + # even if some of the preceding steps are slow. |
| 22 | + timeout-minutes: 25 |
| 23 | + strategy: |
| 24 | + fail-fast: false |
| 25 | + matrix: |
| 26 | + os: |
| 27 | + - macos-latest |
| 28 | + - windows-2022 |
| 29 | + steps: |
| 30 | + - name: Harden Runner |
| 31 | + uses: step-security/harden-runner@0634a2670c59f64b4a01f0f96f84700a4088b9f0 # v2.12.0 |
| 32 | + with: |
| 33 | + egress-policy: audit |
| 34 | + |
| 35 | + # macOS indexes all new files in the background. Our Postgres tests |
| 36 | + # create and destroy thousands of databases on disk, and Spotlight |
| 37 | + # tries to index all of them, seriously slowing down the tests. |
| 38 | + - name: Disable Spotlight Indexing |
| 39 | + if: runner.os == 'macOS' |
| 40 | + run: | |
| 41 | + sudo mdutil -a -i off |
| 42 | + sudo mdutil -X / |
| 43 | + sudo launchctl bootout system /System/Library/LaunchDaemons/com.apple.metadata.mds.plist |
| 44 | +
|
| 45 | + # Set up RAM disks to speed up the rest of the job. This action is in |
| 46 | + # a separate repository to allow its use before actions/checkout. |
| 47 | + - name: Setup RAM Disks |
| 48 | + if: runner.os == 'Windows' |
| 49 | + uses: coder/setup-ramdisk-action@79dacfe70c47ad6d6c0dd7f45412368802641439 |
| 50 | + |
| 51 | + - name: Checkout |
| 52 | + uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2 |
| 53 | + with: |
| 54 | + fetch-depth: 1 |
| 55 | + |
| 56 | + - name: Setup Go |
| 57 | + uses: ./.github/actions/setup-go |
| 58 | + with: |
| 59 | + # Runners have Go baked-in and Go will automatically |
| 60 | + # download the toolchain configured in go.mod, so we don't |
| 61 | + # need to reinstall it. It's faster on Windows runners. |
| 62 | + use-preinstalled-go: ${{ runner.os == 'Windows' }} |
| 63 | + use-temp-cache-dirs: ${{ runner.os == 'Windows' }} |
| 64 | + |
| 65 | + - name: Setup Terraform |
| 66 | + uses: ./.github/actions/setup-tf |
| 67 | + |
| 68 | + - name: Test with PostgreSQL Database |
| 69 | + env: |
| 70 | + POSTGRES_VERSION: "13" |
| 71 | + TS_DEBUG_DISCO: "true" |
| 72 | + LC_CTYPE: "en_US.UTF-8" |
| 73 | + LC_ALL: "en_US.UTF-8" |
| 74 | + shell: bash |
| 75 | + run: | |
| 76 | + if [ "${{ runner.os }}" == "Windows" ]; then |
| 77 | + # Create a temp dir on the R: ramdisk drive for Windows. The default |
| 78 | + # C: drive is extremely slow: https://github.com/actions/runner-images/issues/8755 |
| 79 | + mkdir -p "R:/temp/embedded-pg" |
| 80 | + go run scripts/embedded-pg/main.go -path "R:/temp/embedded-pg" |
| 81 | + fi |
| 82 | + if [ "${{ runner.os }}" == "macOS" ]; then |
| 83 | + # Postgres runs faster on a ramdisk on macOS too |
| 84 | + mkdir -p /tmp/tmpfs |
| 85 | + sudo mount_tmpfs -o noowners -s 8g /tmp/tmpfs |
| 86 | + go run scripts/embedded-pg/main.go -path /tmp/tmpfs/embedded-pg |
| 87 | + fi |
| 88 | +
|
| 89 | + # if macOS, install google-chrome for scaletests |
| 90 | + # As another concern, should we really have this kind of external dependency |
| 91 | + # requirement on standard CI? |
| 92 | + if [ "${{ matrix.os }}" == "macos-latest" ]; then |
| 93 | + brew install google-chrome |
| 94 | + fi |
| 95 | +
|
| 96 | + # By default Go will use the number of logical CPUs, which |
| 97 | + # is a fine default. |
| 98 | + PARALLEL_FLAG="" |
| 99 | +
|
| 100 | + # macOS will output "The default interactive shell is now zsh" |
| 101 | + # intermittently in CI... |
| 102 | + if [ "${{ matrix.os }}" == "macos-latest" ]; then |
| 103 | + touch ~/.bash_profile && echo "export BASH_SILENCE_DEPRECATION_WARNING=1" >> ~/.bash_profile |
| 104 | + fi |
| 105 | +
|
| 106 | + # Golang's default for these 2 variables is the number of logical CPUs. |
| 107 | + # Our Windows and Linux runners have 16 cores, so they match up there. |
| 108 | + NUM_PARALLEL_PACKAGES=16 |
| 109 | + NUM_PARALLEL_TESTS=16 |
| 110 | + if [ "${{ runner.os }}" == "Windows" ]; then |
| 111 | + # On Windows Postgres chokes up when we have 16x16=256 tests |
| 112 | + # running in parallel, and dbtestutil.NewDB starts to take more than |
| 113 | + # 10s to complete sometimes causing test timeouts. With 16x8=128 tests |
| 114 | + # Postgres tends not to choke. |
| 115 | + NUM_PARALLEL_PACKAGES=8 |
| 116 | + fi |
| 117 | + if [ "${{ runner.os }}" == "macOS" ]; then |
| 118 | + # Our macOS runners have 8 cores. We leave NUM_PARALLEL_TESTS at 16 |
| 119 | + # because the tests complete faster and Postgres doesn't choke. It seems |
| 120 | + # that macOS's tmpfs is faster than the one on Windows. |
| 121 | + NUM_PARALLEL_PACKAGES=8 |
| 122 | + fi |
| 123 | +
|
| 124 | + # We rerun failing tests to counteract flakiness coming from Postgres |
| 125 | + # choking on macOS and Windows sometimes. |
| 126 | + DB=ci gotestsum --rerun-fails=2 --rerun-fails-max-failures=1000 \ |
| 127 | + --format standard-quiet --packages "./..." \ |
| 128 | + -- -v -p $NUM_PARALLEL_PACKAGES -parallel=$NUM_PARALLEL_TESTS -count=1 |
| 129 | +
|
| 130 | + - name: Upload test stats to Datadog |
| 131 | + timeout-minutes: 1 |
| 132 | + continue-on-error: true |
| 133 | + uses: ./.github/actions/upload-datadog |
| 134 | + if: success() || failure() |
| 135 | + with: |
| 136 | + api-key: ${{ secrets.DATADOG_API_KEY }} |
| 137 | + |
| 138 | + notify-slack-on-failure: |
| 139 | + needs: |
| 140 | + - test-go-pg |
| 141 | + runs-on: ubuntu-latest |
| 142 | + if: failure() && github.ref == 'refs/heads/main' |
| 143 | + |
| 144 | + steps: |
| 145 | + - name: Send Slack notification |
| 146 | + run: | |
| 147 | + curl -X POST -H 'Content-type: application/json' \ |
| 148 | + --data '{ |
| 149 | + "blocks": [ |
| 150 | + { |
| 151 | + "type": "header", |
| 152 | + "text": { |
| 153 | + "type": "plain_text", |
| 154 | + "text": "❌ Nightly gauntlet failed", |
| 155 | + "emoji": true |
| 156 | + } |
| 157 | + }, |
| 158 | + { |
| 159 | + "type": "section", |
| 160 | + "fields": [ |
| 161 | + { |
| 162 | + "type": "mrkdwn", |
| 163 | + "text": "*Workflow:*\n${{ github.workflow }}" |
| 164 | + }, |
| 165 | + { |
| 166 | + "type": "mrkdwn", |
| 167 | + "text": "*Committer:*\n${{ github.actor }}" |
| 168 | + }, |
| 169 | + { |
| 170 | + "type": "mrkdwn", |
| 171 | + "text": "*Commit:*\n${{ github.sha }}" |
| 172 | + } |
| 173 | + ] |
| 174 | + }, |
| 175 | + { |
| 176 | + "type": "section", |
| 177 | + "text": { |
| 178 | + "type": "mrkdwn", |
| 179 | + "text": "*View failure:* <${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}|Click here>" |
| 180 | + } |
| 181 | + } |
| 182 | + ] |
| 183 | + }' ${{ secrets.CI_FAILURE_SLACK_WEBHOOK }} |
0 commit comments