Fix build

* Spark jars should have a dedicated artifact id * Use python 3 instead of python 2. This requires to make /usr/bin/python point to python3 * Add -l option to docker useradd command to prevent a bug using long uid moby/moby#5419 * Remove \\\ in the maven settings.xml file in the container otherwise the variables criteo.repo.username|password are not correctly used and lead to 401 issues when uploading files to nexus * Fix hive dependency. Groupid has to be org.spark-project.hive and artifactid version 1.2.1.spark2 * use set -e in the build script to fail immediatly when an error occur in any command * Spark scala is a profile and cannot be activated with -D java option but rather with -P maven option * Use --no-transfer-progress in maven commands to make output readable * All mvn commands were using \\ instead of \ leading to bad command interpreation by bash * Fix tar command used to build the 'jar-only' tgz * Fix mvn jar:jar deploy:deploy parameters which are not exactly the same as the ones of mvn deploy:deploy-file * Upgrade pip in the venv otherwise pyarrow cannot install * Fix altDeploymentRepository declaration * Set pypandoc version to 1.5 as some functions have been removed in newer versions
criteo-forks · May 16, 2023 · 632109a · 632109a
1 parent 3b23a78
commit 632109a
Show file tree

Hide file tree

Showing 5 changed files with 61 additions and 60 deletions.
diff --git a/external/docker/criteo-build/Dockerfile b/external/docker/criteo-build/Dockerfile
@@ -57,8 +57,8 @@ RUN yum install -y \
         java-1.8.0-openjdk \
         pinentry-curses \
         pkgconfig \
-        python2-pip \
-        python2-virtualenv \
+        python3-pip \
+        python3-virtualenv \
         rsync \
         ShellCheck \
         sudo \
@@ -86,11 +86,14 @@ ENV MAVEN_OPTS -Xss64m -Xmx2g -XX:ReservedCodeCacheSize=1g
 
 
 RUN groupadd --non-unique -g ${GROUP_ID} ${USER_NAME}
-RUN useradd -g ${GROUP_ID} -u ${USER_ID} -k /root -m ${USER_NAME}
+RUN useradd -l -g ${GROUP_ID} -u ${USER_ID} -k /root -m ${USER_NAME}
 RUN echo "${USER_NAME} ALL=NOPASSWD: ALL" > "/etc/sudoers.d/spark-build-${USER_ID}"
 ENV HOME /home/${USER_NAME}
 RUN mkdir /home/${USER_NAME}/.m2 && chown ${USER_NAME}: /home/${USER_NAME}/.m2
-RUN echo '<settings><mirrors><mirror><id>criteo</id><mirrorOf>*</mirrorOf><url>http://nexus.criteo.prod/content/groups/criteodev</url></mirror></mirrors><servers><server><id>criteo</id><username>\\\${criteo.repo.username}</username><password>\\\${criteo.repo.password}</password></server></servers></settings>' > /home/${USER_NAME}/.m2/settings.xml
+RUN echo '<settings><mirrors><mirror><id>criteo</id><mirrorOf>*</mirrorOf><url>http://nexus.criteo.prod/content/groups/criteodev</url></mirror></mirrors><servers><server><id>criteo</id><username>${criteo.repo.username}</username><password>${criteo.repo.password}</password></server></servers></settings>' > /home/${USER_NAME}/.m2/settings.xml
 
+# Alias python3 to python otherwise python 2 is called
+RUN mv /usr/bin/python /usr/bin/python2
+RUN ln -s /usr/bin/python3 /usr/bin/python
 
 RUN rm -f /var/log/faillog /var/log/lastlog
diff --git a/external/docker/criteo-build/build_config.sh b/external/docker/criteo-build/build_config.sh
@@ -1,2 +1,2 @@
 HDP_VERSION=3.3.0-criteo-20230320100819
-HIVE_VERSION=1.2.1
+HIVE_VERSION=1.2.1.spark2
diff --git a/external/docker/criteo-build/build_script.sh b/external/docker/criteo-build/build_script.sh
@@ -1,4 +1,5 @@
 set -x
+set -e
 
 MAVEN_USER=$1
 MAVEN_PASSWORD=$2
@@ -20,7 +21,7 @@ if [ ${SCALA_RELEASE} == "2.12" ]; then
     MVN_SCALA_PROPERTY="-Pscala-2.12"
 elif [ ${SCALA_RELEASE} == "2.11" ]; then
     ./dev/change-scala-version.sh 2.11
-    MVN_SCALA_PROPERTY="-Dscala-2.11"
+    MVN_SCALA_PROPERTY="-Pscala-2.11"
 else
     echo "[ERROR] Scala release not provided"
     exit 1
@@ -36,83 +37,80 @@ MVN_HDP_ARTIFACT_VERSION="${MVN_ARTIFACT_VERSION}-${HDP_VERSION}"
 SHUFFLE_SERVICE_JAR_FILE="dist/yarn/spark-${CRITEO_VERSION}-yarn-shuffle.jar"
 MVN_COMMON_PROPERTIES="-Dhive.version=${HIVE_VERSION} ${MVN_SCALA_PROPERTY}"
 MVN_COMMON_PROPERTIES_NO_TESTS="${MVN_COMMON_PROPERTIES} -DskipTests"
-MVN_COMMON_NEXUS_PROPERTIES="-DrepositoryId=criteo -Dcriteo.repo.username=${MAVEN_USER} -Dcriteo.repo.password=${MAVEN_PASSWORD} -DretryFailedDeploymentCount=3"
+MVN_COMMON_DEPLOY_FILE_PROPERTIES="-Durl=${NEXUS_ARTIFACT_URL} -DrepositoryId=criteo -Dcriteo.repo.username=${MAVEN_USER} -Dcriteo.repo.password=${MAVEN_PASSWORD} -DretryFailedDeploymentCount=3"
 
 # do some house cleaning
-mvn clean
+mvn --no-transfer-progress clean
 rm -f spark-*.tgz
 rm -f dist/python/dist/*
 rm -f python/dist/*
 
 # change version
-mvn versions:set -DnewVersion=${CRITEO_VERSION}
+mvn --no-transfer-progress versions:set -DnewVersion=${CRITEO_VERSION}
 
 # Build distribution with hadoop
-./dev/make-distribution.sh --pip --name ${SCALA_RELEASE}-${HDP_VERSION} --tgz -Phive -Phive-thriftserver -Pyarn -Dhadoop.version=${HDP_VERSION} ${MVN_COMMON_PROPERTIES_NO_TESTS}
-
+./dev/make-distribution.sh --pip --name ${SCALA_RELEASE}-${HDP_VERSION} --tgz --no-transfer-progress -Phive -Phive-thriftserver -Pyarn -Dhadoop.version=${HDP_VERSION} ${MVN_COMMON_PROPERTIES_NO_TESTS}
 # tgz artifact deployment
-mvn deploy:deploy-file \\
-    --batch-mode \\
-    -DgroupId=com.criteo.tarballs \\
-    -DartifactId=spark \\
-    -Dversion=${MVN_HDP_ARTIFACT_VERSION} \\
-    -Dpackaging=tar.gz \\
-    -Dfile=${SPARK_HDP_ARTIFACT_FILE} \\
-    -Durl=${NEXUS_ARTIFACT_URL} \\
-    ${MVN_COMMON_NEXUS_PROPERTIES}
+mvn deploy:deploy-file \
+    --batch-mode \
+    -DgroupId=com.criteo.tarballs \
+    -DartifactId=spark \
+    -Dversion=${MVN_HDP_ARTIFACT_VERSION} \
+    -Dpackaging=tar.gz \
+    -Dfile=${SPARK_HDP_ARTIFACT_FILE} \
+    ${MVN_COMMON_DEPLOY_FILE_PROPERTIES}
 
 # Build distribution without hadoop
-./dev/make-distribution.sh --pip --name ${SCALA_RELEASE} --tgz -Phive -Phive-thriftserver -Pyarn -Phadoop-provided ${MVN_COMMON_PROPERTIES}
+./dev/make-distribution.sh --pip --name ${SCALA_RELEASE} --tgz --no-transfer-progress -Phive -Phive-thriftserver -Pyarn -Phadoop-provided ${MVN_COMMON_PROPERTIES}
 # tgz artifact deployment
-mvn deploy:deploy-file \\
-    --batch-mode \\
-    -DgroupId=com.criteo.tarballs \\
-    -DartifactId=spark \\
-    -Dversion=${MVN_ARTIFACT_VERSION} \\
-    -Dpackaging=tar.gz \\
-    -Dfile=${SPARK_ARTIFACT_FILE} \\
-    -Durl=${NEXUS_ARTIFACT_URL} \\
-    ${MVN_COMMON_NEXUS_PROPERTIES}
+mvn deploy:deploy-file \
+    --batch-mode \
+    -DgroupId=com.criteo.tarballs \
+    -DartifactId=spark \
+    -Dversion=${MVN_ARTIFACT_VERSION} \
+    -Dpackaging=tar.gz \
+    -Dfile=${SPARK_ARTIFACT_FILE} \
+    ${MVN_COMMON_DEPLOY_FILE_PROPERTIES}
 
 # Create archive with jars only
-cd dist/jars && tar -czf ${OLDPWD}/${SPARK_JARS_ARTIFACT_FILE} dist/jars; cd $OLDPWD
+cd dist/jars && tar -czf ${OLDPWD}/${SPARK_JARS_ARTIFACT_FILE} *.jar; cd $OLDPWD
 
 # Deploy tgz jars only artifact
-mvn deploy:deploy-file \\
-    --batch-mode \\
-    -DgroupId=com.criteo.tarballs \\
-    -DartifactId=spark \\
-    -Dversion=${MVN_ARTIFACT_VERSION} \\
-    -Dpackaging=tar.gz \\
-    -Dfile=${SPARK_JARS_ARTIFACT_FILE} \\
-    -Durl=${NEXUS_ARTIFACT_URL} \\
-    ${MVN_COMMON_NEXUS_PROPERTIES}
+mvn deploy:deploy-file \
+    --batch-mode \
+    -DgroupId=com.criteo.tarballs \
+    -DartifactId=spark-jars \
+    -Dversion=${MVN_ARTIFACT_VERSION} \
+    -Dpackaging=tar.gz \
+    -Dfile=${SPARK_JARS_ARTIFACT_FILE} \
+    ${MVN_COMMON_DEPLOY_FILE_PROPERTIES}
 
 # shuffle service deployment
-mvn deploy:deploy-file \\
-    --batch-mode \\
-    -DgroupId=org.apache.spark \\
-    -DartifactId=yarn-shuffle_${SCALA_RELEASE} \\
-    -Dversion=${CRITEO_VERSION} \\
-    -Dpackaging=jar \\
-    -Dfile=${SHUFFLE_SERVICE_JAR_FILE} \\
-    -Durl=${NEXUS_ARTIFACT_URL} \\
-    ${MVN_COMMON_NEXUS_PROPERTIES}
+mvn deploy:deploy-file \
+    --batch-mode \
+    -DgroupId=org.apache.spark \
+    -DartifactId=yarn-shuffle_${SCALA_RELEASE} \
+    -Dversion=${CRITEO_VERSION} \
+    -Dpackaging=jar \
+    -Dfile=${SHUFFLE_SERVICE_JAR_FILE} \
+    ${MVN_COMMON_DEPLOY_FILE_PROPERTIES}
 
 # jar artifacts (for parent poms) deployment
-mvn jar:jar deploy:deploy \\
-    --batch-mode \\
-    -Phive -Phive-thriftserver \\
-    -Pyarn \\
-    -Phadoop-provided \\
-    -DaltDeploymentRepository=criteo::${NEXUS_ARTIFACT_URL} \\
-    ${MVN_COMMON_NEXUS_PROPERTIES}
+mvn jar:jar deploy:deploy \
+    --batch-mode \
+    -Phive -Phive-thriftserver \
+    -Pyarn \
+    -Phadoop-provided \
+    -DaltDeploymentRepository=criteo::default::${NEXUS_ARTIFACT_URL} \
+    -Dcriteo.repo.username=${MAVEN_USER} \
+    -Dcriteo.repo.password=${MAVEN_PASSWORD}
 
 # python deployment
 pyspark_version=${SPARK_RELEASE}+criteo_${SCALA_RELEASE}.${TIMESTAMP}
 sed -i "s/__version__ = \\\".*\\\"/__version__ = \\\"${pyspark_version}\\\"/g" python/pyspark/version.py
-python2.7 -m venv venv
+python -m venv venv
 source venv/bin/activate
+pip install --upgrade pip
 pip install -r python/requirements.txt
 cd python
 python setup.py bdist_wheel

diff --git a/pom.xml b/pom.xml
@@ -126,7 +126,7 @@
     <flume.version>1.6.0</flume.version>
     <zookeeper.version>3.4.14</zookeeper.version>
     <curator.version>2.13.0</curator.version>
-    <hive.group>org.apache.hive</hive.group>
+    <hive.group>org.spark-project.hive</hive.group>
     <!-- Version used in Maven Hive dependency -->
     <hive.version>1.2.1.spark2</hive.version>
     <!-- Version used for internal directory structure -->
@@ -2632,7 +2632,7 @@
     <profile>
       <id>hive-1.2</id>
       <properties>
-        <hive.group>org.apache.hive</hive.group>
+        <hive.group>org.spark-project.hive</hive.group>
         <hive.classifier></hive.classifier>
         <!-- Version used in Maven Hive dependency -->
         <hive.version>1.2.1.spark2</hive.version>

diff --git a/python/requirements.txt b/python/requirements.txt
@@ -1,7 +1,7 @@
 wheel
 numpy
 pandas
-pypandoc
+pypandoc==1.5
 py4j==0.10.7
 pyarrow
 twine