Fixes and improvements to docker build script and dockerfiles

1) Clean up large Bazel build cache. Total filesystem size reduction as seen by du -sh /:
  devel image: 1.5 GB (Before: 2.9 GB; After: 1.4 GB)
  devel-gpu image: 2.3 GB (Before: 4.7 GB; After: 2.4 GB)
2) Using nvidia-docker for GPU docker build.
3) Upgrade Bazel version from 0.3.1 to 0.3.2.
4) Add missing libcurl3-dev build dependency to devel images.
5) Add scipy and sklearn to Dockerfile.devel-gpu to enhance consistency with other image types (e.g., Dockerfile.devel).
6) Remove the obsolete and unnecessary --recurse-submodules flag for git clone.

Related to GH issues: https://github.com/tensorflow/tensorflow/issues/4116 and https://github.com/tensorflow/tensorflow/issues/4117

However, not using the "git clone --depth 1" suggested by issue #4117, because the size of the git repo is only reduced by about 50 MB by the "--depth 1" flag. This space saving is small compared to the space saving due to bazel cache removal. The complete history of the git repo can be useful for certain development purposes.
Change: 136302103

tensorflow/tools/docker/Dockerfile.devel

@@ -6,6 +6,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
         build-essential \
         curl \
         git \
+        libcurl3-dev \
         libfreetype6-dev \
         libpng12-dev \
         libzmq3-dev \
@@ -65,7 +66,7 @@ RUN echo "build --spawn_strategy=standalone --genrule_strategy=standalone" \
     >>/root/.bazelrc
 ENV BAZELRC /root/.bazelrc
 # Install the most recent bazel release.
-ENV BAZEL_VERSION 0.3.1
+ENV BAZEL_VERSION 0.3.2
 WORKDIR /
 RUN mkdir /bazel && \
     cd /bazel && \
@@ -78,7 +79,7 @@ RUN mkdir /bazel && \
 
 # Download and build TensorFlow.
 
-RUN git clone --recursive https://github.com/tensorflow/tensorflow.git && \
+RUN git clone https://github.com/tensorflow/tensorflow.git && \
     cd tensorflow && \
     git checkout r0.11
 WORKDIR /tensorflow
@@ -87,10 +88,13 @@ WORKDIR /tensorflow
 # more difficult to experiment with local changes. Instead, just add
 # the built directory to the path.
 
-RUN ./configure && \
+RUN tensorflow/tools/ci_build/builds/configured CPU \
     bazel build -c opt tensorflow/tools/pip_package:build_pip_package && \
     bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/pip && \
-    pip install --upgrade /tmp/pip/tensorflow-*.whl
+    pip install --upgrade /tmp/pip/tensorflow-*.whl && \
+    rm -rf /tmp/pip && \
+    rm -rf /root/.cache
+# Clean up pip wheel and Bazel cache when done.
 
 # TensorBoard
 EXPOSE 6006

tensorflow/tools/docker/Dockerfile.devel-gpu

@@ -6,6 +6,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
         build-essential \
         curl \
         git \
+        libcurl3-dev \
         libfreetype6-dev \
         libpng12-dev \
         libzmq3-dev \
@@ -31,6 +32,8 @@ RUN pip --no-cache-dir install \
         jupyter \
         matplotlib \
         numpy \
+        scipy \
+        sklearn \
         && \
     python -m ipykernel.kernelspec
 
@@ -66,7 +69,7 @@ RUN echo "build --spawn_strategy=standalone --genrule_strategy=standalone" \
     >>/root/.bazelrc
 ENV BAZELRC /root/.bazelrc
 # Install the most recent bazel release.
-ENV BAZEL_VERSION 0.3.1
+ENV BAZEL_VERSION 0.3.2
 WORKDIR /
 RUN mkdir /bazel && \
     cd /bazel && \
@@ -79,7 +82,7 @@ RUN mkdir /bazel && \
 
 # Download and build TensorFlow.
 
-RUN git clone -b r0.11 --recursive --recurse-submodules https://github.com/tensorflow/tensorflow.git && \
+RUN git clone https://github.com/tensorflow/tensorflow.git && \
     cd tensorflow && \
     git checkout r0.11
 WORKDIR /tensorflow
@@ -89,10 +92,13 @@ ENV LD_LIBRARY_PATH /usr/local/cuda/extras/CUPTI/lib64:$LD_LIBRARY_PATH
 ENV TF_NEED_CUDA 1
 ENV TF_CUDA_COMPUTE_CAPABILITIES=3.0,3.5,5.2
 
-RUN ./configure && \
+RUN tensorflow/tools/ci_build/builds/configured GPU \
     bazel build -c opt --config=cuda tensorflow/tools/pip_package:build_pip_package && \
     bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/pip && \
-    pip install --upgrade /tmp/pip/tensorflow-*.whl
+    pip install --upgrade /tmp/pip/tensorflow-*.whl && \
+    rm -rf /tmp/pip && \
+    rm -rf /root/.cache
+# Clean up pip wheel and Bazel cache when done.
 
 WORKDIR /root
 

tensorflow/tools/docker/parameterized_docker_build.sh

@@ -120,8 +120,10 @@ else
 fi
 
 if [[ ${TF_DOCKER_BUILD_TYPE} == "cpu" ]]; then
-  :
+  DOCKER_BINARY="docker"
 elif   [[ ${TF_DOCKER_BUILD_TYPE} == "gpu" ]]; then
+  DOCKER_BINARY="nvidia-docker"
+
   FINAL_TAG="${FINAL_TAG}-gpu"
   if [[ ${ORIG_DOCKERFILE} == *"."* ]]; then
     # There is already a dot in the tag, use "-"
@@ -235,17 +237,18 @@ fi
 IMG="${USER}/tensorflow:${FINAL_TAG}"
 echo "Building docker image with image name and tag: ${IMG}"
 
-docker build --no-cache -t "${IMG}" -f "${DOCKERFILE}" "${TMP_DIR}"
+"${DOCKER_BINARY}" build --no-cache -t "${IMG}" -f "${DOCKERFILE}" "${TMP_DIR}"
 if [[ $? == "0" ]]; then
-  echo "docker build of ${IMG} succeeded"
+  echo "${DOCKER_BINARY} build of ${IMG} succeeded"
 else
-  die "FAIL: docker build of ${IMG} with Dockerfile ${DOCKERFILE} failed"
+  die "FAIL: ${DOCKER_BINARY} build of ${IMG} with Dockerfile ${DOCKERFILE} "\
+"failed"
 fi
 
 
 # Make sure that there is no other containers of the same image running
 # TODO(cais): Move to an earlier place.
-if [[ ! -z $(docker ps | grep "${IMG}") ]]; then
+if [[ ! -z $("${DOCKER_BINARY}" ps | grep "${IMG}") ]]; then
   die "ERROR: It appears that there are docker containers of the image "\
 "${IMG} running. Please stop them before proceeding"
 fi
@@ -258,7 +261,7 @@ echo "  (Log file is at: ${DOCKER_RUN_LOG}"
 echo ""
 
 if [[ "${TF_DOCKER_BUILD_IS_DEVEL}" == "no" ]]; then
-  docker run --rm -p ${CONTAINER_PORT}:${CONTAINER_PORT} \
+  "${DOCKER_BINARY}" run --rm -p ${CONTAINER_PORT}:${CONTAINER_PORT} \
       -v ${TMP_DIR}/notebooks:/root/notebooks "${IMG}" \
       2>&1 > "${DOCKER_RUN_LOG}" &
 
@@ -267,7 +270,7 @@ if [[ "${TF_DOCKER_BUILD_IS_DEVEL}" == "no" ]]; then
   while [[ -z ${CONTAINER_ID} ]]; do
     sleep 1
     echo "Polling for container ID..."
-    CONTAINER_ID=$(docker ps | grep "${IMG}" | awk '{print $1}')
+    CONTAINER_ID=$("${DOCKER_BINARY}" ps | grep "${IMG}" | awk '{print $1}')
   done
 
   echo "ID of the running docker container: ${CONTAINER_ID}"
@@ -293,10 +296,10 @@ if [[ "${TF_DOCKER_BUILD_IS_DEVEL}" == "no" ]]; then
 
   # Stop the running docker container
   sleep 1
-  docker stop --time=0 ${CONTAINER_ID}
+  "${DOCKER_BINARY}" stop --time=0 ${CONTAINER_ID}
 
 else
-  docker run --rm -p ${CONTAINER_PORT}:${CONTAINER_PORT} \
+  "${DOCKER_BINARY}" run --rm -p ${CONTAINER_PORT}:${CONTAINER_PORT} \
       -v ${TMP_DIR}/notebooks:/root/notebooks "${IMG}" \
       bash -c \
       "cd /tensorflow; tensorflow/tools/ci_build/builds/test_tutorials.sh"
@@ -324,9 +327,9 @@ fi
 # Apply the final image name and tag
 FINAL_IMG="${FINAL_IMAGE_NAME}:${FINAL_TAG}"
 
-DOCKER_VER=$(docker version | grep Version | head -1 | awk '{print $NF}')
+DOCKER_VER=$("${DOCKER_BINARY}" version | grep Version | head -1 | awk '{print $NF}')
 if [[ -z "${DOCKER_VER}" ]]; then
-  die "ERROR: Failed to determine docker version"
+  die "ERROR: Failed to determine ${DOCKER_BINARY} version"
 fi
 DOCKER_MAJOR_VER=$(echo "${DOCKER_VER}" | cut -d. -f 1)
 DOCKER_MINOR_VER=$(echo "${DOCKER_VER}" | cut -d. -f 2)
@@ -337,7 +340,7 @@ if [[ "${DOCKER_MAJOR_VER}" -le 1 ]] && \
   FORCE_TAG="--force"
 fi
 
-docker tag ${FORCE_TAG} "${IMG}" "${FINAL_IMG}" || \
+"${DOCKER_BINARY}" tag ${FORCE_TAG} "${IMG}" "${FINAL_IMG}" || \
     die "Failed to tag intermediate docker image ${IMG} as ${FINAL_IMG}"
 
 echo ""