diff --git a/elephantscale/docker/Dockerfile b/elephantscale/docker/Dockerfile
index 0e9a549..a7e32ef 100644
--- a/elephantscale/docker/Dockerfile
+++ b/elephantscale/docker/Dockerfile
@@ -1,7 +1,7 @@
 ## pick a specific version, to ensure predictability
-FROM jupyter/pyspark-notebook@sha256:64420e4c348ab48fb806f42332109cbc205ae74cda67240c3e0974c5f7e6e969
+FROM jupyter/pyspark-notebook@sha256:f5f1b391d3f6e94ae864690db1c5e0e9ffecf2f4a0a00f09c01c3e3a9230638e
 ## or latest
-#FROM jupyter/pyspark-notebook@latest
+#FROM jupyter/pyspark-notebook:latest
 
 MAINTAINER Elephant Scale <info@elephantscale.com>
 
@@ -13,10 +13,9 @@ ARG SBT_VERSION=1.0.2
 ARG INSTALL_DIR=/usr/local
 ENV BIGDL_HOME  ${INSTALL_DIR}/BigDL
 
-#ARG BIGDL_URL=https://s3.amazonaws.com/elephantscale-public/BigDL/BigDL.zip
 ## Download BigDL from release page
-## https://bigdl-project.github.io/0.3.0/#release-download/
-ARG BIGDL_URL=https://repo1.maven.org/maven2/com/intel/analytics/bigdl/dist-spark-${SPARK_VERSION}-scala-2.11.8-linux64/0.3.0/dist-spark-${SPARK_VERSION}-scala-2.11.8-linux64-0.3.0-dist.zip
+## https://repo1.maven.org/maven2/com/intel/analytics/bigdl/
+ARG BIGDL_URL=https://repo1.maven.org/maven2/com/intel/analytics/bigdl/dist-spark-2.2.0-scala-2.11.8-all/0.4.0/dist-spark-2.2.0-scala-2.11.8-all-0.4.0-dist.zip
 ## --- end CONFIG
 
 USER root
@@ -30,7 +29,7 @@ RUN apt-get install -yq  --no-install-recommends \
     atop \
     curl \
     less \
-    openjdk-8-jdk-headless \
+    openjdk-8-jdk \
     rsync \
     unzip \
     wget \
@@ -76,6 +75,7 @@ RUN apt-get clean && \
     rm -rf /var/lib/apt/lists/*
 
 ## ----- install BigDL
+RUN echo "BIGDL_URL=$BIGDL_URL"
 RUN \
     mkdir -p ${BIGDL_HOME} && \
     cd ${BIGDL_HOME} && \
@@ -96,15 +96,15 @@ RUN mkdir /work
 USER $NB_USER
 
 ## update conda
-# RUN conda update --all
+#RUN conda update --all
 ## install NLTK
 RUN  conda  install -y  nltk
 
 ## python 3.5 env
-RUN conda create -y -n py35 python=3.5 numpy scipy pandas scikit-learn matplotlib seaborn jupyter nltk tensorflow opencv pillow
+RUN conda create -y -n py35 python=3.5 numpy scipy pandas scikit-learn matplotlib seaborn jupyter nltk opencv pillow tensorflow
 
 ## python 2.7 env
-RUN conda create -y -n py27 python=2.7 numpy scipy pandas scikit-learn matplotlib seaborn jupyter nltk tensorflow opencv pillow
+RUN conda create -y -n py27 python=2.7 numpy scipy pandas scikit-learn matplotlib seaborn jupyter nltk opencv pillow tensorflow
 
 # list envs
 RUN conda info -e
diff --git a/elephantscale/run-bigdl-docker.sh b/elephantscale/run-bigdl-docker.sh
index 6d01f68..28f39e3 100755
--- a/elephantscale/run-bigdl-docker.sh
+++ b/elephantscale/run-bigdl-docker.sh
@@ -3,7 +3,7 @@
 ## Usage
 #  ./run-bigdl-docker.sh   <image name>  [optional command]
 #
-#  ./run-bigdl-docker.sh  xxx/yyy
+#  ./run-bigdl-docker.sh  elephantscale/bigdl-sandbox
 #
 #  or during developing, give a local docker image id
 #  ./run-bigdl-docker.sh  abcd1234
@@ -19,16 +19,12 @@ fi
 
 image_id="$1"
 cmd="$2"
-name="bigdl"
-
-## remove any previously running containers
-docker rm -f "$name"
 
 # mount the current directory at /work
 this="${BASH_SOURCE-$0}"
 mydir=$(cd -P -- "$(dirname -- "$this")" && pwd -P)
 
-docker run -it  --name "$name" \
+docker run -it   \
     -p 8888:8888 \
     -p 6006:6006 \
     -v"$mydir:/work" \