DSP-5892-2: General Shell Script Safety Improvements

Doublequoting most variables to ensure we aren't bitten by variable
expansion.

Author: RussellSpitzer

bin/kill-process-tree.sh

@@ -0,0 +1,60 @@
+#!/bin/bash
+
+if [ "$(which pgrep)" == "" ]; then
+    echo "pgrep is not available" >&2
+    exit 1
+fi
+
+if [ "$#" != "2" ]; then
+    echo "Syntax error, the expected usage is: kill-process-tree <signal> <process_id>" >&2
+    exit 1
+fi
+
+# This simple script allows to kill the whole process tree. The first argument is a signal and the second
+# is the root process
+parent=$2
+signal=$1
+
+# A string which will be populated with descendant processes ids
+pids=""
+
+# Kills the given process if it is running. It also checks whether the given parameter contains process ID
+function kill_if_running() {
+    if [[ "$1" =~ [0-9]+ ]]; then
+        ps -e -o pid= | grep $1 > /dev/null
+        if [ $? -eq 0 ]; then
+            kill "-$signal" "$1"
+            return 0
+        else
+            return 1
+        fi
+    fi
+}
+
+# Populates pids with descendant processes ids
+function get_descendants_pids() {
+    for cpid in $(pgrep -P $1);
+    do
+        pids="$cpid $pids"
+        get_descendants_pids $cpid
+    done
+}
+
+get_descendants_pids $parent
+
+for pid in $pids
+do
+    if [ "$signal" == "" ]; then
+        echo "$pid"
+    else
+        kill_if_running $pid
+        if [ $? -eq 0 ]; then
+            # If the process has been really killed, then wait for 3 seconds before going to the next
+            # process. In most cases, where there is a scripts execution hierarchy with Java process
+            # at the bottom, it is sufficient to kill that Java process and the scripts will just exit
+            # in a natural way
+            echo "Sent $signal to $pid, sleeping for 3 seconds"
+            sleep 3s
+        fi
+    fi
+done

bin/server_package.sh

@@ -44,8 +44,9 @@ fi
 FILES="job-server-extras/target/scala-$majorVersion/spark-job-server.jar
        bin/server_start.sh
        bin/server_stop.sh
+       bin/kill-process-tree.sh
        $CONFIG_DIR/$ENV.conf
-       config/log4j-server.properties"
+       config/logback-server.xml"
 
 rm -rf $WORK_DIR
 mkdir -p $WORK_DIR

bin/server_start.sh

@@ -6,7 +6,7 @@ set -e
 
 get_abs_script_path() {
   pushd . >/dev/null
-  cd $(dirname $0)
+  cd "$(dirname "$0")"
   appdir=$(pwd)
   popd  >/dev/null
 }
@@ -26,14 +26,14 @@ JAVA_OPTS="-XX:MaxDirectMemorySize=512M
 
 MAIN="spark.jobserver.JobServer"
 
-conffile=$(ls -1 $appdir/*.conf | head -1)
+conffile="$(ls -1 "$appdir"/*.conf | head -1)"
 if [ -z "$conffile" ]; then
   echo "No configuration file found"
   exit 1
 fi
 
 if [ -f "$appdir/settings.sh" ]; then
-  . $appdir/settings.sh
+  . "$appdir/settings.sh"
 else
   echo "Missing $appdir/settings.sh, exiting"
   exit 1
@@ -46,7 +46,7 @@ fi
 
 pidFilePath=$appdir/$PIDFILE
 
-if [ -f "$pidFilePath" ] && kill -0 $(cat "$pidFilePath"); then
+if [ -f "$pidFilePath" ] && kill -0 "$(cat "$pidFilePath")"; then
    echo 'Job server is already running'
    exit 1
 fi
@@ -57,8 +57,9 @@ if [ -z "$LOG_DIR" ]; then
 fi
 mkdir -p $LOG_DIR
 
-LOGGING_OPTS="-Dlog4j.configuration=file:$appdir/log4j-server.properties
-              -DLOG_DIR=$LOG_DIR"
+LOGGING_OPTS="-DLOG_DIR=$LOG_DIR"
+
+export SPARK_SUBMIT_LOGBACK_CONF_FILE="$appdir/logback-server.xml"
 
 # For Mesos
 CONFIG_OVERRIDES=""
@@ -78,8 +79,8 @@ fi
 export SPARK_HOME
 
 # DSE_BIN is set in settings.sh
-$DSE_HOME/bin/dse spark-submit --class $MAIN --driver-memory 5G \
+"$DSE_HOME/bin/dse" spark-submit --class "$MAIN" --driver-memory 5G \
   --conf "spark.executor.extraJavaOptions=$LOGGING_OPTS" \
   --driver-java-options "$GC_OPTS $JAVA_OPTS $LOGGING_OPTS $CONFIG_OVERRIDES" \
-  $@ $appdir/spark-job-server.jar $conffile 2>&1 &
-echo $! > $pidFilePath
+  "$@" "$appdir/spark-job-server.jar" "$conffile" 2>&1 &
+echo "$!" > "$pidFilePath"

bin/server_stop.sh

@@ -3,27 +3,30 @@
 
 get_abs_script_path() {
   pushd . >/dev/null
-  cd $(dirname $0)
+  cd "$(dirname "$0")"
   appdir=$(pwd)
   popd  >/dev/null
 }
 
 get_abs_script_path
 
 if [ -f "$appdir/settings.sh" ]; then
-  . $appdir/settings.sh
+  . "$appdir/settings.sh"
 else
   echo "Missing $appdir/settings.sh, exiting"
   exit 1
 fi
 
 pidFilePath=$appdir/$PIDFILE
 
-if [ ! -f "$pidFilePath" ] || ! kill -0 $(cat "$pidFilePath"); then
+if [ ! -f "$pidFilePath" ] || ! kill -0 "$(cat "$pidFilePath")"; then
    echo 'Job server not running'
 else
   echo 'Stopping job server...'
-  PID=$(cat "$pidFilePath")
-  kill -15 -- -$(ps -o pgid= $PID | grep -o [0-9]*) && rm -f "$pidFilePath"
+  PID="$(cat "$pidFilePath")"
+  "$(dirname "$0")"/kill-process-tree.sh -2 $PID
   echo '...job server stopped'
 fi
+
+
+

job-server/config/dse.sh

@@ -10,7 +10,7 @@ APP_GROUP=cassandra
 #Package location for dse-env.sh to get environment variables
 if [ -z "$DSE_ENV" ]; then
     for include in "$HOME/.dse-env.sh" \
-                   "`dirname "$0"`/../../bin/dse-env.sh" \
+                   "`dirname "$0"`/../../../bin/dse-env.sh" \
                    "/etc/dse/dse-env.sh"; do
         if [ -r "$include" ]; then
             DSE_ENV="$include"
@@ -31,15 +31,15 @@ elif [ -z "$ENV" ]; then
     exit 1
 fi
 
-SPARK_VERSION=1.3.1
+SPARK_VERSION=1.3.1.0 #Last digit is DSE Specific 
 
 DEPLOY_HOSTS="localhost"
 
-INSTALL_DIR=$DSE_COMPONENTS_ROOT/spark-jobserver
-LOG_DIR=/var/log/job-server
+INSTALL_DIR="$DSE_COMPONENTS_ROOT/spark/spark-jobserver"
+LOG_DIR=/var/log/spark/job-server
 
 PIDFILE=spark-jobserver.pid
 
-SPARK_CONF_DIR=$SPARK_HOME/conf
+SPARK_CONF_DIR=${SPARK_CONF_DIR:-"$SPARK_HOME/conf"}
 
-SCALA_VERSION=2.10.4 # or 2.11.6
+SCALA_VERSION=2.10.5 # or 2.11.6

version.sbt

@@ -1 +1 @@
-version in ThisBuild := "0.5.2-DSP-5892-SNAPSHOT"
+version in ThisBuild := "0.5.2.DSP-5892-2-SNAPSHOT"