Speed up docker image building and switch base image to alpine

.dockerignore

@@ -20,7 +20,14 @@
 **/*.jar
 **/*.class
 dist
-target
+
+**/target
+
+# To support building the docker image from the distribution file built on host machine,
+# we need to keep binary files under the target directory of distribution module.
+# This rule MUST be after above '**/target' rule
+!distribution/target/*-bin.tar.gz
+
 *.iml
 *.ipr
 *.iws
@@ -35,3 +42,4 @@ target
 *.DS_Store
 _site
 dependency-reduced-pom.xml
+**/node_modules

.github/workflows/distribution-checks.yml

@@ -22,6 +22,7 @@ on:
       - '[0-9]+.[0-9]+.[0-9]+-[A-Za-z0-9]+' # release branches
     paths:
       - 'distribution/**'
+      - 'web-console/**'
       - '**/pom.xml'
   pull_request:
     branches:

distribution/docker/Dockerfile

@@ -19,68 +19,96 @@
 
 ARG JDK_VERSION=17
 
+# By default we build the entire distribution from source in docker
+# This can be unset if the tarball was already built outside of Docker
+ARG BUILD_FROM_SOURCE="true"
+
+#===============================================================
+# web-console building stage
+#===============================================================
 # The platform is explicitly specified as x64 to build the Druid distribution.
 # This is because it's not able to build the distribution on arm64 due to dependency problem of web-console. See: https://github.com/apache/druid/issues/13012
 # Since only java jars are shipped in the final image, it's OK to build the distribution on x64.
 # Once the web-console dependency problem is resolved, we can remove the --platform directive.
-FROM --platform=linux/amd64 maven:3.9 as builder
-
-# Rebuild from source in this stage
-# This can be unset if the tarball was already built outside of Docker
-ARG BUILD_FROM_SOURCE="true"
-
-RUN export DEBIAN_FRONTEND=noninteractive \
-    && apt-get -qq update \
+FROM --platform=linux/amd64 node:20.9.0-slim AS web-console-builder
+
+RUN npm install -g [email protected]
+
+# Only copy necessary files to web-console to build to try the best to use docker cache
+COPY ./web-console /src/web-console
+COPY ./docs /src/docs
+WORKDIR /src/web-console
+
+# Re-declare the ARG to make it available in this stage
+ARG BUILD_FROM_SOURCE
+
+# Suppress a building error saying error:0308010C:digital envelope routines::unsupported
+ENV NODE_OPTIONS=--openssl-legacy-provider
+
+# The directory of output, during the package stage, files will be copied from this directory
+RUN mkdir /output
+
+# The ls command below is just for debug purpose helping us know which files are updated that docker cache are invalidated
+RUN if [ "$BUILD_FROM_SOURCE" = "true" ]; then \
+  ls -l && \
+  npm ci \
+; fi
+RUN if [ "$BUILD_FROM_SOURCE" = "true" ]; then \
+   ./script/build \
+   && ./script/cp-to ./output \
+; fi
+
+#===============================================================
+# Declare a builder stage to build the Druid distribution
+#===============================================================
+FROM maven:3.9.9-eclipse-temurin-${JDK_VERSION} AS builder
+RUN apt-get -qq update \
     && apt-get -qq -y install --no-install-recommends python3 python3-yaml
 
 COPY . /src
 WORKDIR /src
+
+# Copy the web-console resources from the web-console-builder to include it in this stage to package
+COPY --from=web-console-builder /output /src/web-console/target/resources
+
+# Re-declare the ARG to make it available in this stage
+ARG BUILD_FROM_SOURCE
+
+# Note: use web.console.skip to exclude building of web-console but include the package of web-console
 RUN --mount=type=cache,target=/root/.m2 if [ "$BUILD_FROM_SOURCE" = "true" ]; then \
-      mvn -B -ff -q \
+      rm -fr distribution/target/ && \
+      mvn -B -ff -ntp \
       install \
       -Pdist,bundle-contrib-exts \
       -Pskip-static-checks,skip-tests \
-      -Dmaven.javadoc.skip=true -T1C \
-      ; fi
-
-RUN --mount=type=cache,target=/root/.m2 VERSION=$(mvn -B -q org.apache.maven.plugins:maven-help-plugin:3.2.0:evaluate \
-      -Dexpression=project.version -DforceStdout=true \
-    ) \
- && tar -zxf ./distribution/target/apache-druid-${VERSION}-bin.tar.gz -C /opt \
- && mv /opt/apache-druid-${VERSION} /opt/druid
-
-FROM alpine:3 as bash-static
-ARG TARGETARCH
-#
-# Download bash-static binary to execute scripts that require bash.
-# Although bash-static supports multiple platforms, but there's no need for us to support all those platform, amd64 and arm64 are enough.
-#
-ARG BASH_URL_BASE="https://github.com/robxu9/bash-static/releases/download/5.1.016-1.2.3"
-RUN if [ "$TARGETARCH" = "arm64" ]; then \
-      BASH_URL="${BASH_URL_BASE}/bash-linux-aarch64" ; \
-    elif [ "$TARGETARCH" = "amd64" ]; then \
-      BASH_URL="${BASH_URL_BASE}/bash-linux-x86_64" ; \
-    else \
-      echo "Unsupported architecture ($TARGETARCH)" && exit 1; \
-    fi; \
-    echo "Downloading bash-static from ${BASH_URL}" \
-    && wget ${BASH_URL} -O /bin/bash
-
-FROM busybox:1.35.0-glibc as busybox
-
-FROM gcr.io/distroless/java$JDK_VERSION-debian12
+      -Dweb.console.skip=true \
+      -pl '!org.apache.druid:druid-benchmarks, !org.apache.druid:druid-integration-tests, !org.apache.druid.integration-tests:druid-it-tools, !org.apache.druid.integration-tests:druid-it-image, !org.apache.druid.integration-tests:druid-it-cases' \
+      -T1C \
+; fi
+
+
+# RUN ls ./distribution/target
+RUN DISTRIBUTION=$(ls ./distribution/target/apache-druid-*-bin.tar.gz) \
+ && tar -zxf ${DISTRIBUTION} -C /opt \
+ && mv /opt/apache-druid-* /opt/druid
+
+#===============================================================
+# Final stage
+#===============================================================
+# Above bulding staging uses images based on eclipse-temurin,
+# however here we directly use alpine because eclipse-temurin-alpine for JRE17 does not ship a image with ARM64 architecture.
+FROM alpine:latest
 LABEL maintainer="Apache Druid Developers <[email protected]>"
 
-COPY --from=busybox /bin/busybox /busybox/busybox
-RUN ["/busybox/busybox", "--install", "/bin"]
+# Re-daclare the ARG to make it available in this stage
+ARG JDK_VERSION
 
-
-RUN addgroup -S -g 1000 druid \
+# Install necessary packages to easier debugging
+RUN apk add --no-cache openjdk${JDK_VERSION}-jre bash curl net-tools strace lsof \
+ && addgroup -S -g 1000 druid \
  && adduser -S -u 1000 -D -H -h /opt/druid -s /bin/sh -g '' -G druid druid
 
-
-COPY --from=bash-static /bin/bash /bin/bash
-RUN chmod 755 /bin/bash
+ENV JAVA_HOME=/usr/lib/jvm/java-${JDK_VERSION}-openjdk
 
 COPY distribution/docker/druid.sh /druid.sh
 COPY distribution/docker/peon.sh /peon.sh

distribution/docker/README.md

@@ -19,23 +19,18 @@
 
 ## Build
 
-From the root of the repo, run following command:
+From the root of the repo, run the following command:
 
 ```bash
 DOCKER_BUILDKIT=1 docker build -t apache/druid:tag -f distribution/docker/Dockerfile .
 ```
 
-### Building images on Apple M1/M2
-To build images on Apple M1/M2, you need to follow the instructions in this section.
-
-1. build Druid distribution from the root of the repo
-   ```bash
-   mvn clean package -DskipTests -Pdist
-   ```
-2. build target image
-   ```
-   DOCKER_BUILDKIT=1 docker build -t apache/druid:tag -f distribution/docker/Dockerfile --build-arg BUILD_FROM_SOURCE=false .
-   ```
+There are two extra build arguments that can be passed to the build command by using `--build-arg` flag.
+
+| Building argument | Description                                                                                                                                                                       | Default value |
+|-------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------|
+| JDK_VERSION       | The version of JDK to use. By default JDK 17 is used. Use 17 and above.                                                                                                           | 17            |
+| BUILD_FROM_SOURCE | Whether to build Druid from source inside docker.<br/> If you already build the distribution outside the docker, you can manully set this argument to false to speed up building. | true          |
 
 ## Run
 

distribution/docker/druid.sh

@@ -141,7 +141,13 @@ fi
 DRUID_SET_HOST_IP=${DRUID_SET_HOST_IP:-0}
 if [ "${DRUID_SET_HOST_IP}" = "1" ]
 then
-    setKey $SERVICE druid.host $(ip r get 1 | awk '{print $7;exit}')
+    # Get local ip from the 'ip' tool
+    IP=$(ip r get 1 | awk '{print $7;exit}')
+    if [ -z "$IP" ]; then
+      echo "ERROR: IP address not found."
+      exit 1
+    fi
+    setKey $SERVICE druid.host $IP
 fi
 
 env | grep ^druid_ | while read evar;

web-console/package.json

@@ -31,7 +31,7 @@
     "sasslint-fix-changed-only": "npm run sasslint-changed-only -- --fix",
     "prettify": "prettier --write '{src,e2e-tests}/**/*.{ts,tsx,scss}' './*.js'",
     "prettify-check": "prettier --check '{src,e2e-tests}/**/*.{ts,tsx,scss}' './*.js'",
-    "generate-licenses-file": "license-checker --production --json --out licenses.json",
+    "generate-licenses-file": "license-checker --production --json --out target/licenses.json",
     "check-licenses": "license-checker --production --onlyAllow 'Apache-1.1;Apache-2.0;BSD-2-Clause;BSD-3-Clause;0BSD;MIT;ISC;CC0-1.0;OFL-1.1' --summary",
     "start": "webpack serve"
   },

web-console/pom.xml

@@ -34,6 +34,10 @@
   <properties>
     <resources.directory>${project.build.directory}/resources</resources.directory>
     <web.console.skip>false</web.console.skip>  <!-- this property is overidden in Travis CI to skip the javascript-related work -->
+
+    <!--
+    NOTE: When changing the following properties, the docker base image specified in distribution/docker/Dockerfile must be updated accordingly
+    -->
     <node.version>v20.9.0</node.version>
     <npm.version>10.9.0</npm.version>
   </properties>
@@ -150,11 +154,13 @@
         </executions>
       </plugin>
       <plugin>
+        <!--
+        The web.console.skip is NOT applied to this stage as this is required for web-console package when building docker image
+        -->
         <groupId>org.apache.maven.plugins</groupId>
         <artifactId>maven-resources-plugin</artifactId>
         <configuration>
           <outputDirectory>${project.build.outputDirectory}/org/apache/druid/console</outputDirectory>
-          <skip>${web.console.skip}</skip>
         </configuration>
       </plugin>
     </plugins>