Fix: AZTK_IS_MASTER not set on worker and failing

aztk/node_scripts/install/install.py

@@ -29,9 +29,13 @@ def setup_host(docker_repo: str):
     master_node = config.batch_client.compute_node.get(config.pool_id, master_node_id)
 
     if is_master:
-        os.environ["AZTK_IS_MASTER"] = "1"
+        os.environ["AZTK_IS_MASTER"] = "true"
+    else:
+        os.environ["AZTK_IS_MASTER"] = "false"
     if is_worker:
-        os.environ["AZTK_IS_WORKER"] = "1"
+        os.environ["AZTK_IS_WORKER"] = "true"
+    else:
+        os.environ["AZTK_IS_WORKER"] = "false"
 
     os.environ["AZTK_MASTER_IP"] = master_node.ip_address
 
@@ -49,8 +53,8 @@ def setup_spark_container():
     """
     Code run in the main spark container
     """
-    is_master = os.environ["AZTK_IS_MASTER"]
-    is_worker = os.environ["AZTK_IS_WORKER"]
+    is_master = os.environ.get("AZTK_IS_MASTER") == "true"
+    is_worker = os.environ.get("AZTK_IS_WORKER") == "true"
     print("Setting spark container. Master: ", is_master, ", Worker: ", is_worker)
 
     print("Copying spark setup config")

aztk/spark/models/plugins/hdfs/hdfs.sh

@@ -59,7 +59,7 @@ echo '<?xml version="1.0" encoding="UTF-8"?>
     </configuration>' > $HADOOP_HOME/etc/hadoop/hdfs-site.xml
 
 # run HDFS
-if [ "$AZTK_IS_MASTER" -eq "1" ]; then
+if [ "$AZTK_IS_MASTER" -eq "true" ]; then
     echo 'starting namenode and datanode'
     hdfs namenode -format
     $HADOOP_HOME/sbin/hadoop-daemon.sh --config $HADOOP_CONF_DIR --script hdfs start namenode

aztk/spark/models/plugins/jupyter/jupyter.sh

@@ -10,7 +10,7 @@
 
 echo "Is master: $AZTK_IS_MASTER"
 
-if  [ "$AZTK_IS_MASTER" = "1" ]; then
+if  [ "$AZTK_IS_MASTER" = "true" ]; then
     pip install jupyter --upgrade
     pip install notebook --upgrade
 

aztk/spark/models/plugins/jupyter_lab/configuration.py

@@ -15,7 +15,7 @@ def __init__(self):
                     public=True,
                 ),
             ],
-            run_on=PluginTargetRole.All,
+            target_role=PluginTargetRole.All,
             execute="jupyter_lab.sh",
             files=[
                 PluginFile("jupyter_lab.sh", os.path.join(dir_path, "jupyter_lab.sh")),

aztk/spark/models/plugins/jupyter_lab/jupyter_lab.sh

@@ -6,7 +6,7 @@
 #  - aztk/python:spark2.1.0-python3.6.2-base
 #  - aztk/python:spark2.1.0-python3.6.2-gpu
 
-if  [ "$IS_MASTER" = "1" ]; then
+if  [ "$AZTK_IS_MASTER" = "true" ]; then
     conda install -c conda-force jupyterlab
 
     PYSPARK_DRIVER_PYTHON="/.pyenv/versions/${USER_PYTHON_VERSION}/bin/jupyter"

aztk/spark/models/plugins/rstudio_server/rstudio_server.sh

@@ -7,7 +7,7 @@
 #  - jiata/aztk-r:0.1.0-spark2.1.0-r3.4.1
 #  - jiata/aztk-r:0.1.0-spark1.6.3-r3.4.1
 
-if  [ "$AZTK_IS_MASTER" = "1" ]; then
+if  [ "$AZTK_IS_MASTER" = "true" ]; then
 
     ## Download and install Rstudio Server
     wget https://download2.rstudio.org/rstudio-server-$RSTUDIO_SERVER_VERSION-amd64.deb

custom-scripts/hdfs.sh

@@ -59,7 +59,7 @@ echo '<?xml version="1.0" encoding="UTF-8"?>
     </configuration>' > $HADOOP_HOME/etc/hadoop/hdfs-site.xml
 
 # run HDFS
-if [ "$AZTK_IS_MASTER" -eq "1" ]; then
+if [ "$AZTK_IS_MASTER" -eq "true" ]; then
     echo 'starting namenode and datanode'
     hdfs namenode -format
     $HADOOP_HOME/sbin/hadoop-daemon.sh --config $HADOOP_CONF_DIR --script hdfs start namenode

custom-scripts/jupyter.sh

@@ -8,7 +8,7 @@
 #  - aztk/python:spark2.1.0-python3.6.2-base
 #  - aztk/python:spark2.1.0-python3.6.2-gpu
 
-if  [ "$AZTK_IS_MASTER" = "1" ]; then
+if  [ "$AZTK_IS_MASTER" = "true" ]; then
     pip install jupyter --upgrade
     pip install notebook --upgrade
 

custom-scripts/rstudio_server.sh

@@ -1,13 +1,13 @@
 #!/bin/bash
 
 # This custom script only works on images where rstudio server is pre-installed on the Docker image
-# 
+#
 # This custom script has been tested to work on the following docker images:
 #  - jiata/aztk-r:0.1.0-spark2.2.0-r3.4.1
 #  - jiata/aztk-r:0.1.0-spark2.1.0-r3.4.1
 #  - jiata/aztk-r:0.1.0-spark1.6.3-r3.4.1
 
-if  [ "$AZTK_IS_MASTER" = "1" ]; then
+if  [ "$AZTK_IS_MASTER" = "true" ]; then
 
     ## Download and install Rstudio Server
     wget https://download2.rstudio.org/rstudio-server-$RSTUDIO_SERVER_VERSION-amd64.deb
@@ -19,7 +19,7 @@ if  [ "$AZTK_IS_MASTER" = "1" ]; then
     set -e
     useradd -m -d /home/rstudio rstudio -g staff
     echo rstudio:rstudio | chpasswd
-    
+
     rstudio-server start
 
 fi

custom-scripts/simple.sh

@@ -2,11 +2,11 @@
 
 # Demo plugin. Not actually meant to be used.
 
-if  [ "$AZTK_IS_MASTER" = "1" ]; then
+if  [ "$AZTK_IS_MASTER" = "true" ]; then
     echo "This is a custom script running on just the master!"
 fi
 
-if  [ "$AZTK_IS_WORKER" = "1" ]; then
+if  [ "$AZTK_IS_WORKER" = "true" ]; then
     echo "This is a custom script running on just the workers!"
 fi
 

docs/51-define-plugin.md

@@ -70,8 +70,8 @@ cluster_config = ClusterConfiguration(
 
 AZTK provide a few environment variables that can be used in your plugin script
 
-* `AZTK_IS_MASTER`: Is the plugin running on the master node
-* `AZTK_IS_WORKER`: Is a worker setup on the current node(This might also be a master if you have `worker_on_master` set to true)
+* `AZTK_IS_MASTER`: Is the plugin running on the master node. Can be either `true` or `false`
+* `AZTK_IS_WORKER`: Is a worker setup on the current node(This might also be a master if you have `worker_on_master` set to true) Can be either `true` or `false`
 * `AZTK_MASTER_IP`: Internal ip of the master
 
 ## Debug your plugin