[docker] Build docker in Travis PR & Fix image build failing (#9787)

Co-authored-by: Ian Rodney <ian.rodney@gmail.com>

.travis.yml

@@ -126,6 +126,7 @@ matrix:
         - . ./ci/travis/ci.sh build
       script:
         - . ./ci/travis/ci.sh test_wheels
+        - if [[ "$TRAVIS_PULL_REQUEST" != "false" ]]; then ./ci/keep_alive bash $TRAVIS_BUILD_DIR/ci/travis/build-docker-images.sh; fi
         - bash ./java/build-jar-multiplatform.sh linux
         - bash ./java/build-jar-multiplatform.sh multiplatform
         #- bash ./java/build-jar-multiplatform.sh deploy
@@ -431,7 +432,7 @@ deploy:
 
   - provider: script
     edge: true # This supposedly opts in to deploy v2.
-    script: ./ci/keep_alive bash $TRAVIS_BUILD_DIR/ci/travis/build-autoscaler-images.sh || true
+    script: ./ci/keep_alive bash $TRAVIS_BUILD_DIR/ci/travis/build-docker-images.sh
     skip_cleanup: true
     on:
       repo: ray-project/ray

ci/travis/build-autoscaler-images.sh

@@ -1,48 +0,0 @@
-#!/usr/bin/env bash
-
-# This script build docker images for autoscaler.
-# For now, we only build python3.6 images.
-set -e
-set -x
-
-SCRIPT_DIR=$(cd "$(dirname "${BASH_SOURCE:-$0}")"; pwd)
-ROOT_DIR=$(cd "$SCRIPT_DIR"/../../; pwd)
-DOCKER_USERNAME="raytravisbot"
-
-# We will only build and push when we are building branch build.
-if [[ "$TRAVIS" == "true" && "$TRAVIS_PULL_REQUEST" == "false" ]]; then
-    echo "$DOCKER_PASSWORD" | docker login -u "$DOCKER_USERNAME" --password-stdin
-
-    wheel="$(basename "$ROOT_DIR"/.whl/*cp36m-manylinux*)"
-    commit_sha=$(echo "$TRAVIS_COMMIT" | head -c 6)
-    cp -r "$ROOT_DIR"/.whl "$ROOT_DIR"/docker/autoscaler/.whl
-
-    docker build -q -t rayproject/base-deps docker/base-deps
-
-    docker build \
-        --build-arg WHEEL_PATH=".whl/$wheel" \
-        --build-arg WHEEL_NAME="$wheel" \
-        -t rayproject/autoscaler:"$commit_sha" \
-        "$ROOT_DIR"/docker/autoscaler
-
-    docker tag rayproject/base-deps rayproject/base-deps:"$commit_sha" 
-    docker push rayproject/base-deps:"$commit_sha"
-    docker push rayproject/autoscaler:"$commit_sha"
-
-
-    # We have a branch build, e.g. release/v0.7.0
-    if [[ "$TRAVIS_BRANCH" != "master" ]]; then
-       # Replace / in branch name to - so it is legal tag name
-       normalized_branch_name=$(echo "$TRAVIS_BRANCH" | sed -e "s/\//-/")
-       docker tag rayproject/autoscaler:"$commit_sha" rayproject/autoscaler:"$normalized_branch_name"
-       docker tag rayproject/base-deps:"$commit_sha" rayproject/base-deps:"$normalized_branch_name"
-       docker push rayproject/autoscaler:"$normalized_branch_name"
-       docker push rayproject/base-deps:"$normalized_branch_name"
-    else
-       docker tag rayproject/autoscaler:"$commit_sha" rayproject/autoscaler:latest
-       docker tag rayproject/base-deps:"$commit_sha" rayproject/base-deps:latest
-       docker push rayproject/autoscaler:latest
-       docker push rayproject/base-deps:latest
-    fi
-fi
-

ci/travis/build-docker-images.sh

@@ -0,0 +1,75 @@
+#!/usr/bin/env bash
+
+# This script build docker images for autoscaler.
+# For now, we only build python3.6 images.
+set -e
+set -x
+
+SCRIPT_DIR=$(cd "$(dirname "${BASH_SOURCE:-$0}")"; pwd)
+ROOT_DIR=$(cd "$SCRIPT_DIR"/../../; pwd)
+DOCKER_USERNAME="raytravisbot"
+
+docker_push() {
+    if [[ "$TRAVIS_PULL_REQUEST" == "false" ]]; then
+        docker push $@
+    else
+        echo "Skipping docker push because it's in PR environment."
+    fi
+}
+
+# We will only build and push when we are building branch build.
+if [[ "$TRAVIS" == "true" ]]; then
+
+    if [[ "$TRAVIS_PULL_REQUEST" == "false" ]]; then
+        echo "$DOCKER_PASSWORD" | docker login -u "$DOCKER_USERNAME" --password-stdin
+    else
+        if [[ "$RAY_CI_DOCKER_AFFECTED" == "0" ]]; then
+            echo "Skipping docker build in PR build because dockerfile didn't change."
+            exit 0
+        fi
+    fi
+
+    wheel="$(basename "$ROOT_DIR"/.whl/*cp37m-manylinux*)"
+    commit_sha=$(echo "$TRAVIS_COMMIT" | head -c 6)
+    cp -r "$ROOT_DIR"/.whl "$ROOT_DIR"/docker/ray/.whl
+    cp $ROOT_DIR/python/requirements.txt $ROOT_DIR/docker/autoscaler/requirements.txt
+    cp $ROOT_DIR/python/requirements_autoscaler.txt $ROOT_DIR/docker/autoscaler/requirements_autoscaler.txt
+
+    docker build -t rayproject/base-deps docker/base-deps
+
+    docker build \
+        --build-arg WHEEL_PATH=".whl/$wheel" \
+        -t rayproject/ray \
+        "$ROOT_DIR"/docker/ray
+
+    docker build \
+        -t rayproject/autoscaler:"$commit_sha" \
+        "$ROOT_DIR"/docker/autoscaler
+
+    docker tag rayproject/base-deps rayproject/base-deps:"$commit_sha" 
+    docker tag rayproject/ray rayproject/ray:"$commit_sha" 
+    docker_push rayproject/base-deps:"$commit_sha"
+    docker_push rayproject/ray:"$commit_sha"
+    docker_push rayproject/autoscaler:"$commit_sha"
+
+
+    # We have a branch build, e.g. release/v0.7.0
+    if [[ "$TRAVIS_BRANCH" != "master" ]]; then
+       # Replace / in branch name to - so it is legal tag name
+       normalized_branch_name=$(echo "$TRAVIS_BRANCH" | sed -e "s/\//-/")
+       docker tag rayproject/autoscaler:"$commit_sha" rayproject/autoscaler:"$normalized_branch_name"
+       docker tag rayproject/ray:"$commit_sha" rayproject/ray:"$normalized_branch_name"
+       docker tag rayproject/base-deps:"$commit_sha" rayproject/base-deps:"$normalized_branch_name"
+       docker_push rayproject/autoscaler:"$normalized_branch_name"
+       docker_push rayproject/ray:"$normalized_branch_name"
+       docker_push rayproject/base-deps:"$normalized_branch_name"
+    else
+       docker tag rayproject/autoscaler:"$commit_sha" rayproject/autoscaler:latest
+       docker tag rayproject/ray:"$commit_sha" rayproject/ray:latest
+       docker tag rayproject/base-deps:"$commit_sha" rayproject/base-deps:latest
+       docker_push rayproject/autoscaler:latest
+       docker_push rayproject/ray:latest
+       docker_push rayproject/base-deps:latest
+    fi
+fi
+

ci/travis/ci.sh

@@ -287,7 +287,9 @@ build_wheels() {
 
       # This command should be kept in sync with ray/python/README-building-wheels.md,
       # except the "${MOUNT_BAZEL_CACHE[@]}" part.
+      # TODO(ilr) Re-add "suppress_output" to show info
       docker run --rm -w /ray -v "${PWD}":/ray "${MOUNT_BAZEL_CACHE[@]}" \
+        -e TRAVIS_COMMIT="${TRAVIS_COMMIT}"  -e CI="${CI}" \
         rayproject/arrow_linux_x86_64_base:python-3.8.0 /ray/python/build-wheel-manylinux1.sh
       ;;
     darwin*)

ci/travis/determine_tests_to_run.py

@@ -46,6 +46,7 @@ if __name__ == "__main__":
     RAY_CI_STREAMING_PYTHON_AFFECTED = 0
     RAY_CI_STREAMING_JAVA_AFFECTED = 0
     RAY_CI_DASHBOARD_AFFECTED = 0
+    RAY_CI_DOCKER_AFFECTED = 0
 
     event_type = None
     for key in ["GITHUB_EVENT_NAME", "TRAVIS_EVENT_TYPE"]:
@@ -102,6 +103,8 @@ if __name__ == "__main__":
             elif changed_file.startswith("java/"):
                 RAY_CI_JAVA_AFFECTED = 1
                 RAY_CI_STREAMING_JAVA_AFFECTED = 1
+            elif changed_file.startswith("docker/"):
+                RAY_CI_DOCKER_AFFECTED = 1
             elif any(
                     changed_file.startswith(prefix)
                     for prefix in skip_prefix_list):

docker/autoscaler/Dockerfile

@@ -1,6 +1,4 @@
-FROM rayproject/base-deps
-ARG WHEEL_PATH
-ARG WHEEL_NAME
+FROM rayproject/ray
 
 # Install kubectl.
 RUN apt-get update \
@@ -13,14 +11,10 @@ RUN apt-get update \
     && apt-get clean 
 
 # We have to uninstall wrapt this way for Tensorflow compatibility
-COPY ../../python/requirements.txt ./
-COPY ../../python/requirements_autoscaler.txt ./
-COPY $WHEEL_PATH $WHEEL_NAME
+COPY requirements.txt ./
+COPY requirements_autoscaler.txt ./
 
-RUN pip --no-cache-dir install -r requirements.txt \ 
-    && pip --no-cache-dir install -r requirements_autoscaler.txt \
-    && pip --no-cache-dir install $WHEEL_NAME[all]
+RUN $HOME/anaconda3/bin/pip --no-cache-dir install -r requirements.txt \ 
+    && $HOME/anaconda3/bin/pip --no-cache-dir install -r requirements_autoscaler.txt \
+    && rm requirements.txt && rm requirements_autoscaler.txt
 
-# For Click
-ENV LC_ALL=C.UTF-8
-ENV LANG=C.UTF-8

docker/base-deps/Dockerfile

@@ -1,6 +1,8 @@
 # The base-deps Docker image installs main libraries needed to run Ray
 
 FROM ubuntu:focal
+ENV TZ=America/Los_Angeles
+ARG DEBIAN_FRONTEND=noninteractive
 RUN apt-get update -y && apt-get upgrade -y \
     && apt-get install -y \
         git \
@@ -9,13 +11,11 @@ RUN apt-get update -y && apt-get upgrade -y \
         build-essential \
         curl \
         unzip \
-        libgtk2.0-dev \
-        zlib1g-dev \
-        libgl1-mesa-dev \
         tmux \
         screen \
         rsync \
         openssh-client \
+    && rm -rf /var/lib/apt/lists/* \
     && apt-get clean \
     && wget \
         --quiet "https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh" \
@@ -35,4 +35,5 @@ RUN apt-get update -y && apt-get upgrade -y \
     # AttributeError: 'numpy.ufunc' object has no attribute '__module__'
     && $HOME/anaconda3/bin/pip uninstall -y dask
 
-ENV PATH "$HOME/anaconda3/bin:$PATH"
+# TODO(ilr) $HOME seems to point to result in "" instead of "/root"
+ENV PATH "/root/anaconda3/bin:$PATH"

docker/examples/Dockerfile

@@ -5,7 +5,7 @@ FROM ray-project/deploy
 # This updates numpy to 1.14 and mutes errors from other libraries
 RUN conda install -y numpy
 # Needed to run Tune example with a 'plot' call - which does not actually render a plot, but throws an error.
-RUN apt-get install -y zlib1g-dev libgl1-mesa-dev
+RUN apt-get install -y zlib1g-dev libgl1-mesa-dev libgtk2.0-dev
 RUN pip install -U pip
 RUN pip install gym[atari] opencv-python-headless tensorflow lz4 pytest-timeout smart_open tensorflow_probability dm_tree
 RUN pip install -U h5py  # Mutes FutureWarnings

docker/ray/Dockerfile

@@ -0,0 +1,7 @@
+FROM rayproject/base-deps
+ARG WHEEL_PATH
+# For Click
+ENV LC_ALL=C.UTF-8
+ENV LANG=C.UTF-8
+COPY $WHEEL_PATH .
+RUN $HOME/anaconda3/bin/pip --no-cache-dir install `basename $WHEEL_PATH`[all]