chore: build boost libs from source

.scripts/upload_new_boost_version.sh

@@ -0,0 +1,59 @@
+#!/usr/bin/env bash
+# This script mirrors the boostorg/boost source bundle for the given version to Nexus.
+# The boost source bundle is architecture independent.
+# It contains its own build system (b2) which is also built from source before building boost itself, so we don't need to worry about architecture specific builds.
+# This artifact is used by the hadoop/boost local image.
+
+
+set -euo pipefail
+
+VERSION=${1:?"Missing version number argument (arg 1)"}
+NEXUS_USER=${2:?"Missing Nexus username argument (arg 2)"}
+
+read -r -s -p "Nexus Password: " NEXUS_PASSWORD
+echo ""
+
+# https://stackoverflow.com/questions/4632028/how-to-create-a-temporary-directory
+# Find the directory name of the script
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
+
+# the temp directory used, within $DIR
+WORK_DIR=$(mktemp -d -p "$DIR")
+
+# check if tmp dir was created
+if [[ ! "$WORK_DIR" || ! -d "$WORK_DIR" ]]; then
+  echo "Could not create temp dir"
+  exit 1
+fi
+
+# deletes the temp directory
+function cleanup {
+  rm -rf "$WORK_DIR"
+}
+
+# register the cleanup function to be called on the EXIT signal
+trap cleanup EXIT
+
+cd "$WORK_DIR" || exit
+
+# boost does not currently publish signatures or SBOMs
+BOOST_UNDERSCORE="$(echo "${VERSION}" | tr '.' '_')"
+BOOST_TARBALL="boost_${BOOST_UNDERSCORE}.tar.bz2"
+DOWNLOAD_URL="https://archives.boost.io/release/$VERSION/source/$BOOST_TARBALL"
+
+echo "Downloading boost"
+if ! curl --fail -Ls -O "$DOWNLOAD_URL"; then
+  echo "Failed to download from $DOWNLOAD_URL"
+  exit 1
+fi
+
+FILE_NAME=$(basename "$DOWNLOAD_URL")
+
+echo "Uploading boost to Nexus"
+if ! curl --fail -o /dev/null --progress-bar -u "$NEXUS_USER:$NEXUS_PASSWORD" --upload-file "$FILE_NAME" 'https://repo.stackable.tech/repository/packages/boost/'; then
+  echo "Failed to upload boost to Nexus"
+  exit 1
+fi
+
+echo "Successfully uploaded new version of boost ($VERSION) to Nexus"
+echo "https://repo.stackable.tech/service/rest/repository/browse/packages/boost/"

CHANGELOG.md

@@ -12,6 +12,8 @@ All notable changes to this project will be documented in this file.
 ### Changed
 
 - hbase: Update `hbase-opa-authorizer` from `0.1.0` to `0.2.0` and then `0.3.0` ([#1446], [#1454]).
+- hadoop: Build the boost library from source instead of installing the EPEL package ([#1475]).
+          This is a requirement for the adoption of UBI 10 as base image because the package is not available there.
 
 ### Fixed
 
@@ -25,6 +27,7 @@ All notable changes to this project will be documented in this file.
 [#1463]: https://github.com/stackabletech/docker-images/pull/1463
 [#1466]: https://github.com/stackabletech/docker-images/pull/1466
 [#1474]: https://github.com/stackabletech/docker-images/pull/1474
+[#1475]: https://github.com/stackabletech/docker-images/pull/1475
 
 ## [26.3.0] - 2026-03-16
 

hadoop/Dockerfile

@@ -90,6 +90,7 @@ LABEL \
 
 COPY --chown=${STACKABLE_USER_UID}:0 --from=hadoop-builder /stackable/hadoop-${HADOOP_VERSION}-stackable${RELEASE_VERSION} /stackable/hadoop-${HADOOP_VERSION}-stackable${RELEASE_VERSION}
 COPY --chown=${STACKABLE_USER_UID}:0 --from=hadoop-builder /stackable/*-src.tar.gz /stackable
+COPY --chown=${STACKABLE_USER_UID}:0 --from=hadoop-builder /stackable/boost /stackable/boost
 
 COPY --chown=${STACKABLE_USER_UID}:0 --from=hdfs-utils-builder /stackable/hdfs-utils-${HDFS_UTILS_VERSION}.jar /stackable/hadoop-${HADOOP_VERSION}-stackable${RELEASE_VERSION}/share/hadoop/common/lib/hdfs-utils-${HDFS_UTILS_VERSION}.jar
 COPY --chown=${STACKABLE_USER_UID}:0 --from=hdfs-utils-builder /stackable/hdfs-utils-${HDFS_UTILS_VERSION}-src.tar.gz /stackable
@@ -105,9 +106,11 @@ microdnf update
 # tar is required for `kubectl cp` which can be used to copy the log files
 # or profiler flamegraph from the Pod
 # It is already installed in the base image but leaving here for documentation purposes
+# libstdc++ is a runtime dependency for boost
 microdnf install \
   fuse \
   fuse-libs \
+  libstdc++ \
   tar
 microdnf clean all
 rpm -qa --qf "%{NAME}-%{VERSION}-%{RELEASE_VERSION}\n" | sort > /stackable/package_manifest.txt
@@ -132,8 +135,13 @@ chmod -x "/stackable/jmx/jmx_prometheus_javaagent-${JMX_EXPORTER_VERSION}.jar"
 ln -s "/stackable/jmx/jmx_prometheus_javaagent-${JMX_EXPORTER_VERSION}.jar" /stackable/jmx/jmx_prometheus_javaagent.jar
 
 # Set correct permissions and ownerships
-chown --recursive ${STACKABLE_USER_UID}:0 /stackable/hadoop /stackable/jmx /stackable/async-profiler "/stackable/async-profiler-${ASYNC_PROFILER_VERSION}-${TARGETOS}-${ARCH}"
-chmod --recursive g=u /stackable/jmx /stackable/async-profiler "/stackable/hadoop-${HADOOP_VERSION}-stackable${RELEASE_VERSION}"
+chown --recursive ${STACKABLE_USER_UID}:0 \
+/stackable/hadoop /stackable/jmx /stackable/async-profiler \
+"/stackable/async-profiler-${ASYNC_PROFILER_VERSION}-${TARGETOS}-${ARCH}" \
+/stackable/boost
+chmod --recursive g=u /stackable/jmx /stackable/async-profiler \
+"/stackable/hadoop-${HADOOP_VERSION}-stackable${RELEASE_VERSION}" \
+/stackable/boost
 
 # Workaround for https://issues.apache.org/jira/browse/HADOOP-12845
 # The problem is that our stackable-devel image does contain the openssl-devel package
@@ -162,7 +170,7 @@ EOF
 USER ${STACKABLE_USER_UID}
 
 ENV HOME=/stackable
-ENV LD_LIBRARY_PATH=/stackable/hadoop/lib/native:/usr/lib/jvm/jre/lib/server
+ENV LD_LIBRARY_PATH=/stackable/boost/lib:/stackable/hadoop/lib/native:/usr/lib/jvm/jre/lib/server
 ENV PATH="${PATH}":/stackable/hadoop/bin
 ENV HADOOP_HOME=/stackable/hadoop
 ENV HADOOP_CONF_DIR=/stackable/config
@@ -174,6 +182,8 @@ ENV ASYNC_PROFILER_HOME=/stackable/async-profiler
 # if HADOOP_YARN_HOME does not exist at all, so we set it here to a sensible default.
 ENV HADOOP_YARN_HOME=/stackable/hadoop
 ENV HADOOP_MAPRED_HOME=/stackable/hadoop
+ENV BOOST_ROOT=/stackable/boost
+ENV CPATH=/stackable/boost/include
 
 WORKDIR /stackable/hadoop
 CMD ["echo", "This image is not meant to be 'run' directly."]

hadoop/boost/Dockerfile

@@ -0,0 +1,40 @@
+# syntax=docker/dockerfile:1.16.0@sha256:e2dd261f92e4b763d789984f6eab84be66ab4f5f08052316d8eb8f173593acf7
+# check=error=true
+
+# The boost library is a runtime dependency of the Hadoop native code
+# but is not available in the RedHat UBI10 or EPEL10 repositories, so we need to build
+# it ourselves.
+#
+# NOTE: We use a published source bundle instead of the patchable workflow
+#       because boost uses git submodules for its build system
+#       and patchable doesn't support these.
+#       The source bundle contains everything needed to build b2 (the build system)
+#       and boost except for dependencies which are automatically discovered.
+#
+FROM local-image/stackable-base AS boost-builder
+
+ARG BOOST_VERSION
+
+WORKDIR /tmp
+
+RUN <<EOF
+microdnf install \
+      gcc gcc-c++ make perl python3 \
+      tar bzip2 xz
+microdnf clean all
+
+BOOST_UNDERSCORE="$(echo "${BOOST_VERSION}" | tr '.' '_')"
+BOOST_TARBALL="boost_${BOOST_UNDERSCORE}.tar.bz2"
+# See .scripts/upload_new_boost_version.sh for how this artifact got there.
+BOOST_URL="https://repo.stackable.tech/repository/packages/boost/${BOOST_TARBALL}"
+
+mkdir -p /stackable/boost
+curl -fsSL "${BOOST_URL}" -o "${BOOST_TARBALL}"
+tar -xf "${BOOST_TARBALL}"
+cd "boost_${BOOST_UNDERSCORE}"
+./bootstrap.sh --prefix=/stackable/boost
+./b2 -j"$(nproc)" \
+  link=shared runtime-link=shared variant=release threading=multi \
+  install
+rm -rf /tmp/*
+EOF

hadoop/boost/boil-config.toml

@@ -0,0 +1,5 @@
+[versions."1.78.0".local-images]
+stackable-base = "1.0.0"
+
+[versions."1.78.0".build-arguments]
+boost-version = "1.78.0"

hadoop/hadoop/Dockerfile

@@ -1,6 +1,7 @@
 # syntax=docker/dockerfile:1.16.0@sha256:e2dd261f92e4b763d789984f6eab84be66ab4f5f08052316d8eb8f173593acf7
 # check=error=true
 
+FROM local-image/hadoop/boost AS boost-builder
 FROM local-image/java-devel AS hadoop-builder
 
 ARG PRODUCT_VERSION
@@ -14,17 +15,21 @@ COPY --chown=${STACKABLE_USER_UID}:0 shared/protobuf/stackable/patches/patchable
 COPY --chown=${STACKABLE_USER_UID}:0 shared/protobuf/stackable/patches/${PROTOBUF_VERSION} /stackable/src/shared/protobuf/stackable/patches/${PROTOBUF_VERSION}
 
 RUN <<EOF
-rpm --install --replacepkgs https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm
-microdnf update
-# boost is a build dependency starting in Hadoop 3.4.0 if compiling native code
-# automake and libtool are required to build protobuf
-microdnf install boost1.78-devel automake libtool
+# libstdc++ is a runtime dependency for boost,
+# automake and libtool are needed to build protobuf
+microdnf install libstdc++ automake libtool
 microdnf clean all
 rm -rf /var/cache/yum
 mkdir /opt/protobuf
 chown ${STACKABLE_USER_UID}:0 /opt/protobuf
 EOF
 
+COPY --chown=${STACKABLE_USER_UID}:0 --from=boost-builder /stackable/boost /stackable/boost
+
+ENV BOOST_ROOT=/stackable/boost
+ENV LD_LIBRARY_PATH=/stackable/boost/lib
+ENV CPATH=/stackable/boost/include
+
 USER ${STACKABLE_USER_UID}
 # This Protobuf version is the exact version as used in the Hadoop Dockerfile
 # See https://github.com/apache/hadoop/blob/trunk/dev-support/docker/pkg-resolver/install-protobuf.sh

hadoop/hadoop/boil-config.toml

@@ -1,11 +1,13 @@
 [versions."3.3.6".local-images]
 java-devel = "11"
+"hadoop/boost" = "1.78.0"
 
 [versions."3.3.6".build-arguments]
 protobuf-version = "3.7.1"
 
 [versions."3.4.2".local-images]
 java-devel = "11"
+"hadoop/boost" = "1.78.0"
 
 [versions."3.4.2".build-arguments]
 protobuf-version = "3.7.1"