Add support for Centos9 Stream + GCC12

.github/workflows/docker.yml

@@ -48,10 +48,10 @@ jobs:
             file: "scripts/check-container.dockfile"
             args: "cpu_target=avx"
             tags: "ghcr.io/facebookincubator/velox-dev:check-avx"
-          - name: Centos 8
+          - name: Centos 9
             file: "scripts/centos.dockerfile"
             args: "cpu_target=avx"
-            tags: "ghcr.io/facebookincubator/velox-dev:centos8"
+            tags: "ghcr.io/facebookincubator/velox-dev:centos9"
           - name: Dev
             file: "scripts/ubuntu-22.04-cpp.dockerfile"
             args: ""

.github/workflows/experimental.yml

@@ -134,7 +134,7 @@ jobs:
       - name: "Build"
         run: |
           cd velox
-          source /opt/rh/gcc-toolset-9/enable
+          source /opt/rh/gcc-toolset-12/enable
           make debug NUM_THREADS="${{ inputs.numThreads || 8 }}" MAX_HIGH_MEM_JOBS="${{ inputs.maxHighMemJobs || 8 }}" MAX_LINK_JOBS="${{ inputs.maxLinkJobs || 4 }}" EXTRA_CMAKE_FLAGS="-DVELOX_ENABLE_ARROW=ON ${{ inputs.extraCMakeFlags }}"
           ccache -s
 

.github/workflows/linux-build.yml

@@ -61,7 +61,7 @@ jobs:
       VELOX_DEPENDENCY_SOURCE: SYSTEM
       simdjson_SOURCE: BUNDLED
       xsimd_SOURCE: BUNDLED
-      CUDA_VERSION: "11.8"
+      CUDA_VERSION: "12.4"
     steps:
       - uses: actions/checkout@v4
 
@@ -70,6 +70,15 @@ jobs:
         # it doesn't work
         run: git config --global --add safe.directory /__w/velox/velox
 
+      - name: Install Dependencies
+        run: |
+          # Allows to install arbitrary cuda-version whithout needing to update
+          # docker container before. It simplifies testing new/different versions
+          if ! yum list installed cuda-nvcc-$(echo ${CUDA_VERSION} | tr '.' '-') 1>/dev/null; then
+            source scripts/setup-centos9.sh
+            install_cuda ${CUDA_VERSION}
+          fi
+
       - uses: assignUser/stash/restore@v1
         with:
           path: '${{ env.CCACHE_DIR }}'
@@ -82,6 +91,10 @@ jobs:
       - name: Make Release Build
         env:
           MAKEFLAGS: 'NUM_THREADS=8 MAX_HIGH_MEM_JOBS=4 MAX_LINK_JOBS=4'
+          CUDA_ARCHITECTURES: 70
+          CUDA_COMPILER: /usr/local/cuda-${CUDA_VERSION}/bin/nvcc
+          # Set compiler to GCC 12
+          CUDA_FLAGS: "-ccbin /opt/rh/gcc-toolset-12/root/usr/bin"
         run: |
           EXTRA_CMAKE_FLAGS=(
             "-DVELOX_ENABLE_BENCHMARKS=ON"
@@ -92,6 +105,7 @@ jobs:
             "-DVELOX_ENABLE_GCS=ON"
             "-DVELOX_ENABLE_ABFS=ON"
             "-DVELOX_ENABLE_REMOTE_FUNCTIONS=ON"
+            "-DVELOX_ENABLE_GPU=ON"
           )
           make release EXTRA_CMAKE_FLAGS="${EXTRA_CMAKE_FLAGS[*]}"
     

.github/workflows/scheduled.yml

@@ -89,7 +89,7 @@ jobs:
     # prevent errors when forks ff their main branch
     if: ${{ github.repository == 'facebookincubator/velox' }}
     runs-on: 16-core-ubuntu
-    container: ghcr.io/facebookincubator/velox-dev:centos8
+    container: ghcr.io/facebookincubator/velox-dev:centos9
     timeout-minutes: 120
     env:
       CCACHE_DIR: "/__w/velox/velox/.ccache"
@@ -323,7 +323,7 @@ jobs:
     name: Presto Fuzzer
     if: ${{ needs.compile.outputs.presto_bias  != 'true' }}
     runs-on: ubuntu-latest
-    container: ghcr.io/facebookincubator/velox-dev:centos8
+    container: ghcr.io/facebookincubator/velox-dev:centos9
     needs: compile
     timeout-minutes: 120
     steps:
@@ -394,7 +394,7 @@ jobs:
   presto-bias-fuzzer:
     name: Presto Bias Fuzzer
     runs-on: ubuntu-latest
-    container: ghcr.io/facebookincubator/velox-dev:centos8
+    container: ghcr.io/facebookincubator/velox-dev:centos9
     needs: compile
     if: ${{ needs.compile.outputs.presto_bias  == 'true' }}
     timeout-minutes: 120
@@ -445,7 +445,7 @@ jobs:
   spark-aggregate-fuzzer-run:
     name: Spark Aggregate Fuzzer
     runs-on: ubuntu-latest
-    container: ghcr.io/facebookincubator/velox-dev:centos8
+    container: ghcr.io/facebookincubator/velox-dev:centos9
     needs: compile
     timeout-minutes: 60
     steps:
@@ -479,7 +479,7 @@ jobs:
   spark-bias-fuzzer:
     name: Spark Bias Fuzzer
     runs-on: ubuntu-latest
-    container: ghcr.io/facebookincubator/velox-dev:centos8
+    container: ghcr.io/facebookincubator/velox-dev:centos9
     needs: compile
     if: ${{ needs.compile.outputs.spark_bias  == 'true' }}
     timeout-minutes: 120
@@ -523,7 +523,7 @@ jobs:
     name: Spark Fuzzer
     if: ${{ needs.compile.outputs.spark_bias  != 'true' }}
     runs-on: ubuntu-latest
-    container: ghcr.io/facebookincubator/velox-dev:centos8
+    container: ghcr.io/facebookincubator/velox-dev:centos9
     needs: compile
     timeout-minutes: 120
     steps:
@@ -564,7 +564,7 @@ jobs:
   join-fuzzer-run:
     name: Join Fuzzer
     runs-on: ubuntu-latest
-    container: ghcr.io/facebookincubator/velox-dev:centos8
+    container: ghcr.io/facebookincubator/velox-dev:centos9
     needs: compile
     timeout-minutes: 120
     steps:
@@ -598,7 +598,7 @@ jobs:
   exchange-fuzzer-run:
     name: Exchange Fuzzer
     runs-on: ubuntu-latest
-    container: ghcr.io/facebookincubator/velox-dev:centos8
+    container: ghcr.io/facebookincubator/velox-dev:centos9
     needs: compile
     timeout-minutes: 120
     steps:
@@ -634,7 +634,7 @@ jobs:
   row-number-fuzzer-run:
     name: RowNumber Fuzzer
     runs-on: ubuntu-latest
-    container: ghcr.io/facebookincubator/velox-dev:centos8
+    container: ghcr.io/facebookincubator/velox-dev:centos9
     needs: compile
     timeout-minutes: 120
     steps:

docker-compose.yml

@@ -33,19 +33,45 @@ services:
       - .:/velox:delegated
     command: scripts/docker-command.sh
 
+  adapters-cpp:
+    # Usage:
+    #   docker-compose pull adapters-cpp or docker-compose build adapters-cpp
+    #   or
+    #   docker-compose run --rm adapters-cpp
+    #   or
+    #   docker-compose run -e NUM_THREADS=<NUMBER_OF_THREADS_TO_USE> --rm adapters-cpp
+    #   to set the number of threads used during compilation
+    # scripts/adapters.dockerfile uses SHELL which is not supported for OCI image format.
+    # podman users must specify "--podman-build-args='--format docker'" argument.
+    image: ghcr.io/facebookincubator/velox-dev:adapters
+    build:
+      context: .
+      dockerfile: scripts/adapters.dockerfile
+      args:
+        image: ghcr.io/facebookincubator/velox-dev:centos9
+    environment:
+      NUM_THREADS: 8 # default value for NUM_THREADS
+      CCACHE_DIR: "/velox/.ccache"
+      EXTRA_CMAKE_FLAGS: -DVELOX_ENABLE_PARQUET=ON 
+                         -DVELOX_ENABLE_S3=ON
+    volumes:
+      - .:/velox:delegated
+    working_dir: /velox
+    command: /velox/scripts/docker-command.sh
+
   centos-cpp:
     # Usage:
     #   docker-compose pull centos-cpp or docker-compose build centos-cpp
     #   docker-compose run --rm centos-cpp
     #   or
     #   docker-compose run -e NUM_THREADS=<NUMBER_OF_THREADS_TO_USE> --rm centos-cpp
     #   to set the number of threads used during compilation
-    image: ghcr.io/facebookincubator/velox-dev:centos8
+    image: ghcr.io/facebookincubator/velox-dev:centos9
     build:
       context: .
       dockerfile: scripts/centos.dockerfile
       args:
-        image: quay.io/centos/centos:stream8 
+        image: quay.io/centos/centos:stream9
     environment:
       NUM_THREADS: 8 # default value for NUM_THREADS
       CCACHE_DIR: "/velox/.ccache"

scripts/adapters.dockerfile

@@ -12,13 +12,13 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # Build the test and build container for presto_cpp
-ARG image=ghcr.io/facebookincubator/velox-dev:centos8
+ARG image=ghcr.io/facebookincubator/velox-dev:centos9
 FROM $image
 ARG cpu_target=avx
 ENV CPU_TARGET=$cpu_target
 
 COPY scripts/setup-adapters.sh /
-RUN mkdir build && ( cd build &&  source /opt/rh/gcc-toolset-9/enable && \
+RUN mkdir build && ( cd build &&  source /opt/rh/gcc-toolset-12/enable && \
     bash /setup-adapters.sh ) && rm -rf build && dnf remove -y conda && dnf clean all
 
 # install miniforge
@@ -41,5 +41,5 @@ ENV HADOOP_HOME=/usr/local/hadoop \
     LIBHDFS3_CONF=/velox/scripts/hdfs-client.xml \
     PATH=/usr/local/hadoop/bin:${PATH}
 
-ENTRYPOINT ["/bin/bash", "-c", "source /opt/rh/gcc-toolset-9/enable && exec \"$@\"", "--"]
+ENTRYPOINT ["/bin/bash", "-c", "source /opt/rh/gcc-toolset-12/enable && exec \"$@\"", "--"]
 CMD ["/bin/bash"]

scripts/centos.dockerfile

@@ -12,25 +12,24 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # Build the test and build container for presto_cpp
-ARG image=quay.io/centos/centos:stream8
+ARG image=quay.io/centos/centos:stream9
 FROM $image
 ARG cpu_target=avx
 ENV CPU_TARGET=$cpu_target
 
 COPY scripts/setup-helper-functions.sh /
-COPY scripts/setup-centos8.sh /
+COPY scripts/setup-centos9.sh /
 # The removal of the build dir has to happen in the same layer as the build
 # to minimize the image size. gh & jq are required for CI
-RUN mkdir build && ( cd build && bash /setup-centos8.sh ) && rm -rf build && \
+RUN mkdir build && ( cd build && bash /setup-centos9.sh ) && rm -rf build && \
         dnf install -y -q 'dnf-command(config-manager)' && \
         dnf config-manager --add-repo 'https://cli.github.com/packages/rpm/gh-cli.repo' && \
-        dnf install -y -q gh jq python39 && \
-        dnf clean all && \
-        alternatives --remove python3 /usr/bin/python3.6
+        dnf install -y -q gh jq && \
+        dnf clean all
 
 
-ENV CC=/opt/rh/gcc-toolset-9/root/bin/gcc \
-    CXX=/opt/rh/gcc-toolset-9/root/bin/g++
+ENV CC=/opt/rh/gcc-toolset-12/root/bin/gcc \
+    CXX=/opt/rh/gcc-toolset-12/root/bin/g++
 
-ENTRYPOINT ["/bin/bash", "-c", "source /opt/rh/gcc-toolset-9/enable && exec \"$@\"", "--"]
+ENTRYPOINT ["/bin/bash", "-c", "source /opt/rh/gcc-toolset-12/enable && exec \"$@\"", "--"]
 CMD ["/bin/bash"]

scripts/prestojava-container.dockerfile

@@ -13,7 +13,7 @@
 # limitations under the License.
 # Build the test and build container for presto_cpp
 #
-FROM ghcr.io/facebookincubator/velox-dev:centos8 
+FROM ghcr.io/facebookincubator/velox-dev:centos9 
 
 ARG PRESTO_VERSION=0.286
 

scripts/setup-adapters.sh

@@ -61,7 +61,7 @@ function install_gcs-sdk-cpp {
   github_checkout abseil/abseil-cpp 20240116.2 --depth 1
   cmake_install \
     -DABSL_BUILD_TESTING=OFF \
-    -DCMAKE_CXX_STANDARD=14 \
+    -DCMAKE_CXX_STANDARD=17 \
     -DABSL_PROPAGATE_CXX_STD=ON \
     -DABSL_ENABLE_INSTALL=ON
 
@@ -169,12 +169,12 @@ if [[ "$OSTYPE" == "linux-gnu"* ]]; then
       # Dependencies of Azure Storage Blob cpp
       apt install -y openssl
    else # Assume Fedora/CentOS
-      yum -y install libxml2-devel libgsasl-devel libuuid-devel
+      dnf -y install libxml2-devel libgsasl-devel libuuid-devel krb5-devel
       # Dependencies of GCS, probably a workaround until the docker image is rebuilt
-      yum -y install curl-devel c-ares-devel
+      dnf -y install npm curl-devel c-ares-devel
       # Dependencies of Azure Storage Blob Cpp
-      yum -y install perl-IPC-Cmd
-      yum -y install openssl
+      dnf -y install perl-IPC-Cmd
+      dnf -y install openssl
    fi
 fi