bin/hbase-daemon.sh

   1 #!/usr/bin/env bash
   2 #
   3 #/**
   4 # * Licensed to the Apache Software Foundation (ASF) under one
   5 # * or more contributor license agreements.  See the NOTICE file
   6 # * distributed with this work for additional information
   7 # * regarding copyright ownership.  The ASF licenses this file
   8 # * to you under the Apache License, Version 2.0 (the
   9 # * "License"); you may not use this file except in compliance
  10 # * with the License.  You may obtain a copy of the License at
  11 # *
  12 # *     http://www.apache.org/licenses/LICENSE-2.0
  13 # *
  14 # * Unless required by applicable law or agreed to in writing, software
  15 # * distributed under the License is distributed on an "AS IS" BASIS,
  16 # * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  17 # * See the License for the specific language governing permissions and
  18 # * limitations under the License.
  19 # */
  20 #
  21 # Runs a Hadoop hbase command as a daemon.
  22 #
  23 # Environment Variables
  24 #
  25 #   HBASE_CONF_DIR   Alternate hbase conf dir. Default is ${HBASE_HOME}/conf.
  26 #   HBASE_LOG_DIR    Where log files are stored.  PWD by default.
  27 #   HBASE_PID_DIR    The pid files are stored. /tmp by default.
  28 #   HBASE_IDENT_STRING   A string representing this instance of hadoop. $USER by default
  29 #   HBASE_NICENESS The scheduling priority for daemons. Defaults to 0.
  30 #   HBASE_STOP_TIMEOUT  Time, in seconds, after which we kill -9 the server if it has not stopped.
  31 #                        Default 1200 seconds.
  32 #
  33 # Modelled after $HADOOP_HOME/bin/hadoop-daemon.sh
  34
  35 usage="Usage: hbase-daemon.sh [--config <conf-dir>]\
  36  [--autostart-window-size <window size in hours>]\
  37  [--autostart-window-retry-limit <retry count limit for autostart>]\
  38  (start|stop|restart|autostart|autorestart|foreground_start) <hbase-command> \
  39  <args...>"
  40
  41 # if no args specified, show usage
  42 if [ $# -le 1 ]; then
  43   echo $usage
  44   exit 1
  45 fi
  46
  47 # default autostart args value indicating infinite window size and no retry limit
  48 AUTOSTART_WINDOW_SIZE=0
  49 AUTOSTART_WINDOW_RETRY_LIMIT=0
  50
  51 bin=`dirname "${BASH_SOURCE-$0}"`
  52 bin=`cd "$bin">/dev/null; pwd`
  53
  54 . "$bin"/hbase-config.sh
  55 . "$bin"/hbase-common.sh
  56
  57 # get arguments
  58 startStop=$1
  59 shift
  60
  61 command=$1
  62 shift
  63
  64 hbase_rotate_log ()
  65 {
  66     log=$1;
  67     num=5;
  68     if [ -n "$2" ]; then
  69     num=$2
  70     fi
  71     if [ -f "$log" ]; then # rotate logs
  72     while [ $num -gt 1 ]; do
  73         prev=`expr $num - 1`
  74         [ -f "$log.$prev" ] && mv -f "$log.$prev" "$log.$num"
  75         num=$prev
  76     done
  77     mv -f "$log" "$log.$num";
  78     fi
  79 }
  80
  81 cleanAfterRun() {
  82   if [ -f ${HBASE_PID} ]; then
  83     # If the process is still running time to tear it down.
  84     kill -9 `cat ${HBASE_PID}` > /dev/null 2>&1
  85     rm -f ${HBASE_PID} > /dev/null 2>&1
  86   fi
  87
  88   if [ -f ${HBASE_ZNODE_FILE} ]; then
  89     if [ "$command" = "master" ]; then
  90       HBASE_OPTS="$HBASE_OPTS $HBASE_MASTER_OPTS" $bin/hbase master clear > /dev/null 2>&1
  91     else
  92       #call ZK to delete the node
  93       ZNODE=`cat ${HBASE_ZNODE_FILE}`
  94       HBASE_OPTS="$HBASE_OPTS $HBASE_REGIONSERVER_OPTS" $bin/hbase zkcli delete ${ZNODE} > /dev/null 2>&1
  95     fi
  96     rm ${HBASE_ZNODE_FILE}
  97   fi
  98 }
  99
 100 check_before_start(){
 101     #ckeck if the process is not running
 102     mkdir -p "$HBASE_PID_DIR"
 103     if [ -f $HBASE_PID ]; then
 104       if kill -0 `cat $HBASE_PID` > /dev/null 2>&1; then
 105         echo $command running as process `cat $HBASE_PID`.  Stop it first.
 106         exit 1
 107       fi
 108     fi
 109 }
 110
 111 wait_until_done ()
 112 {
 113     p=$1
 114     cnt=${HBASE_SLAVE_TIMEOUT:-300}
 115     origcnt=$cnt
 116     while kill -0 $p > /dev/null 2>&1; do
 117       if [ $cnt -gt 1 ]; then
 118         cnt=`expr $cnt - 1`
 119         sleep 1
 120       else
 121         echo "Process did not complete after $origcnt seconds, killing."
 122         kill -9 $p
 123         exit 1
 124       fi
 125     done
 126     return 0
 127 }
 128
 129 # get log directory
 130 if [ "$HBASE_LOG_DIR" = "" ]; then
 131   export HBASE_LOG_DIR="$HBASE_HOME/logs"
 132 fi
 133 mkdir -p "$HBASE_LOG_DIR"
 134
 135 if [ "$HBASE_PID_DIR" = "" ]; then
 136   HBASE_PID_DIR=/tmp
 137 fi
 138
 139 if [ "$HBASE_IDENT_STRING" = "" ]; then
 140   export HBASE_IDENT_STRING="$USER"
 141 fi
 142
 143 # Some variables
 144 # Work out java location so can print version into log.
 145 if [ "$JAVA_HOME" != "" ]; then
 146   #echo "run java in $JAVA_HOME"
 147   JAVA_HOME=$JAVA_HOME
 148 fi
 149 if [ "$JAVA_HOME" = "" ]; then
 150   echo "Error: JAVA_HOME is not set."
 151   exit 1
 152 fi
 153
 154 JAVA=$JAVA_HOME/bin/java
 155 export HBASE_LOG_PREFIX=hbase-$HBASE_IDENT_STRING-$command-$HOSTNAME
 156 export HBASE_LOGFILE=$HBASE_LOG_PREFIX.log
 157
 158 if [ -z "${HBASE_ROOT_LOGGER}" ]; then
 159 export HBASE_ROOT_LOGGER=${HBASE_ROOT_LOGGER:-"INFO,RFA"}
 160 fi
 161
 162 if [ -z "${HBASE_SECURITY_LOGGER}" ]; then
 163 export HBASE_SECURITY_LOGGER=${HBASE_SECURITY_LOGGER:-"INFO,RFAS"}
 164 fi
 165
 166 HBASE_LOGOUT=${HBASE_LOGOUT:-"$HBASE_LOG_DIR/$HBASE_LOG_PREFIX.out"}
 167 HBASE_LOGGC=${HBASE_LOGGC:-"$HBASE_LOG_DIR/$HBASE_LOG_PREFIX.gc"}
 168 HBASE_LOGLOG=${HBASE_LOGLOG:-"${HBASE_LOG_DIR}/${HBASE_LOGFILE}"}
 169 HBASE_PID=$HBASE_PID_DIR/hbase-$HBASE_IDENT_STRING-$command.pid
 170 export HBASE_ZNODE_FILE=$HBASE_PID_DIR/hbase-$HBASE_IDENT_STRING-$command.znode
 171 export HBASE_AUTOSTART_FILE=$HBASE_PID_DIR/hbase-$HBASE_IDENT_STRING-$command.autostart
 172
 173 if [ -n "$SERVER_GC_OPTS" ]; then
 174   export SERVER_GC_OPTS=${SERVER_GC_OPTS/"-Xloggc:<FILE-PATH>"/"-Xloggc:${HBASE_LOGGC}"}
 175 fi
 176 if [ -n "$CLIENT_GC_OPTS" ]; then
 177   export CLIENT_GC_OPTS=${CLIENT_GC_OPTS/"-Xloggc:<FILE-PATH>"/"-Xloggc:${HBASE_LOGGC}"}
 178 fi
 179
 180 # Set default scheduling priority
 181 if [ "$HBASE_NICENESS" = "" ]; then
 182     export HBASE_NICENESS=0
 183 fi
 184
 185 thiscmd="$bin/$(basename ${BASH_SOURCE-$0})"
 186 args=$@
 187
 188 case $startStop in
 189
 190 (start)
 191     check_before_start
 192     hbase_rotate_log $HBASE_LOGOUT
 193     hbase_rotate_log $HBASE_LOGGC
 194     echo running $command, logging to $HBASE_LOGOUT
 195     $thiscmd --config "${HBASE_CONF_DIR}" \
 196         foreground_start $command $args < /dev/null > ${HBASE_LOGOUT} 2>&1  &
 197     disown -h -r
 198     sleep 1; head "${HBASE_LOGOUT}"
 199   ;;
 200
 201 (autostart)
 202     check_before_start
 203     hbase_rotate_log $HBASE_LOGOUT
 204     hbase_rotate_log $HBASE_LOGGC
 205     echo running $command, logging to $HBASE_LOGOUT
 206     nohup $thiscmd --config "${HBASE_CONF_DIR}" --autostart-window-size ${AUTOSTART_WINDOW_SIZE} --autostart-window-retry-limit ${AUTOSTART_WINDOW_RETRY_LIMIT} \
 207         internal_autostart $command $args < /dev/null > ${HBASE_LOGOUT} 2>&1  &
 208   ;;
 209
 210 (autorestart)
 211     echo running $command, logging to $HBASE_LOGOUT
 212     # stop the command
 213     $thiscmd --config "${HBASE_CONF_DIR}" stop $command $args &
 214     wait_until_done $!
 215     # wait a user-specified sleep period
 216     sp=${HBASE_RESTART_SLEEP:-3}
 217     if [ $sp -gt 0 ]; then
 218       sleep $sp
 219     fi
 220
 221     check_before_start
 222     hbase_rotate_log $HBASE_LOGOUT
 223     hbase_rotate_log $HBASE_LOGGC
 224     nohup $thiscmd --config "${HBASE_CONF_DIR}" --autostart-window-size ${AUTOSTART_WINDOW_SIZE} --autostart-window-retry-limit ${AUTOSTART_WINDOW_RETRY_LIMIT} \
 225         internal_autostart $command $args < /dev/null > ${HBASE_LOGOUT} 2>&1  &
 226   ;;
 227
 228 (foreground_start)
 229     trap cleanAfterRun SIGHUP SIGINT SIGTERM EXIT
 230     if [ "$HBASE_NO_REDIRECT_LOG" != "" ]; then
 231         # NO REDIRECT
 232         echo "`date` Starting $command on `hostname`"
 233         echo "`ulimit -a`"
 234         # in case the parent shell gets the kill make sure to trap signals.
 235         # Only one will get called. Either the trap or the flow will go through.
 236         nice -n $HBASE_NICENESS "$HBASE_HOME"/bin/hbase \
 237             --config "${HBASE_CONF_DIR}" \
 238             $command "$@" start &
 239     else
 240         echo "`date` Starting $command on `hostname`" >> ${HBASE_LOGLOG}
 241         echo "`ulimit -a`" >> "$HBASE_LOGLOG" 2>&1
 242         # in case the parent shell gets the kill make sure to trap signals.
 243         # Only one will get called. Either the trap or the flow will go through.
 244         nice -n $HBASE_NICENESS "$HBASE_HOME"/bin/hbase \
 245             --config "${HBASE_CONF_DIR}" \
 246             $command "$@" start >> ${HBASE_LOGOUT} 2>&1 &
 247     fi
 248     # Add to the command log file vital stats on our environment.
 249     hbase_pid=$!
 250     echo $hbase_pid > ${HBASE_PID}
 251     wait $hbase_pid
 252   ;;
 253
 254 (internal_autostart)
 255     ONE_HOUR_IN_SECS=3600
 256     autostartWindowStartDate=`date +%s`
 257     autostartCount=0
 258     touch "$HBASE_AUTOSTART_FILE"
 259
 260     # keep starting the command until asked to stop. Reloop on software crash
 261     while true
 262     do
 263       if [ -f $HBASE_PID ] &&  kill -0 "$(cat "$HBASE_PID")" > /dev/null 2>&1 ; then
 264         wait "$(cat "$HBASE_PID")"
 265       else
 266         #if the file does not exist it means that it was not stopped properly by the stop command
 267         if [ ! -f "$HBASE_AUTOSTART_FILE" ]; then
 268           echo "`date` HBase might be stopped removing the autostart file. Exiting Autostart process" >> ${HBASE_LOGOUT}
 269           exit 1
 270         fi
 271
 272         echo "`date` Autostarting hbase $command service. Attempt no: $(( $autostartCount + 1))" >> ${HBASE_LOGLOG}
 273         touch "$HBASE_AUTOSTART_FILE"
 274         $thiscmd --config "${HBASE_CONF_DIR}" foreground_start $command $args
 275         autostartCount=$(( $autostartCount + 1 ))
 276
 277         # HBASE-6504 - only take the first line of the output in case verbose gc is on
 278         distMode=`$bin/hbase --config "$HBASE_CONF_DIR" org.apache.hadoop.hbase.util.HBaseConfTool hbase.cluster.distributed | head -n 1`
 279
 280         if [ "$distMode" != 'false' ]; then
 281           #if the cluster is being stopped then do not restart it again.
 282           zparent=`$bin/hbase org.apache.hadoop.hbase.util.HBaseConfTool zookeeper.znode.parent`
 283           if [ "$zparent" == "null" ]; then zparent="/hbase"; fi
 284           zkrunning=`$bin/hbase org.apache.hadoop.hbase.util.HBaseConfTool zookeeper.znode.state`
 285           if [ "$zkrunning" == "null" ]; then zkrunning="running"; fi
 286           zkFullRunning=$zparent/$zkrunning
 287           $bin/hbase zkcli stat $zkFullRunning 2>&1 | grep "Node does not exist"  1>/dev/null 2>&1
 288
 289           #grep returns 0 if it found something, 1 otherwise
 290           if [ $? -eq 0 ]; then
 291             echo "`date` hbase znode does not exist. Exiting Autostart process" >> ${HBASE_LOGOUT}
 292             rm -f "$HBASE_AUTOSTART_FILE"
 293             exit 1
 294           fi
 295
 296           #If ZooKeeper cannot be found, then do not restart
 297           $bin/hbase zkcli stat $zkFullRunning 2>&1 | grep Exception | grep ConnectionLoss  1>/dev/null 2>&1
 298           if [ $? -eq 0 ]; then
 299             echo "`date` zookeeper not found. Exiting Autostart process" >> ${HBASE_LOGOUT}
 300             rm -f "$HBASE_AUTOSTART_FILE"
 301             exit 1
 302           fi
 303         fi
 304       fi
 305
 306       curDate=`date +%s`
 307       autostartWindowReset=false
 308
 309       # reset the auto start window size if it exceeds
 310       if [ $AUTOSTART_WINDOW_SIZE -gt 0 ] && [ $(( $curDate - $autostartWindowStartDate )) -gt $(( $AUTOSTART_WINDOW_SIZE * $ONE_HOUR_IN_SECS )) ]; then
 311         echo "Resetting Autorestart window size: $autostartWindowStartDate" >> ${HBASE_LOGOUT}
 312         autostartWindowStartDate=$curDate
 313         autostartWindowReset=true
 314         autostartCount=0
 315       fi
 316
 317       # kill autostart if the retry limit is exceeded within the given window size (window size other then 0)
 318       if ! $autostartWindowReset && [ $AUTOSTART_WINDOW_RETRY_LIMIT -gt 0 ] && [ $autostartCount -gt $AUTOSTART_WINDOW_RETRY_LIMIT ]; then
 319         echo "`date` Autostart window retry limit: $AUTOSTART_WINDOW_RETRY_LIMIT exceeded for given window size: $AUTOSTART_WINDOW_SIZE hours.. Exiting..." >> ${HBASE_LOGLOG}
 320         rm -f "$HBASE_AUTOSTART_FILE"
 321         exit 1
 322       fi
 323
 324       # wait for shutdown hook to complete
 325       sleep 20
 326     done
 327   ;;
 328
 329 (stop)
 330     echo running $command, logging to $HBASE_LOGOUT
 331     rm -f "$HBASE_AUTOSTART_FILE"
 332     if [ -f $HBASE_PID ]; then
 333       pidToKill=`cat $HBASE_PID`
 334       # kill -0 == see if the PID exists
 335       if kill -0 $pidToKill > /dev/null 2>&1; then
 336         echo -n stopping $command
 337         echo "`date` Terminating $command" >> $HBASE_LOGLOG
 338         kill $pidToKill > /dev/null 2>&1
 339         waitForProcessEnd $pidToKill $command
 340       else
 341         retval=$?
 342         echo no $command to stop because kill -0 of pid $pidToKill failed with status $retval
 343       fi
 344     else
 345       echo no $command to stop because no pid file $HBASE_PID
 346     fi
 347     rm -f $HBASE_PID
 348   ;;
 349
 350 (restart)
 351     echo running $command, logging to $HBASE_LOGOUT
 352     # stop the command
 353     $thiscmd --config "${HBASE_CONF_DIR}" stop $command $args &
 354     wait_until_done $!
 355     # wait a user-specified sleep period
 356     sp=${HBASE_RESTART_SLEEP:-3}
 357     if [ $sp -gt 0 ]; then
 358       sleep $sp
 359     fi
 360     # start the command
 361     $thiscmd --config "${HBASE_CONF_DIR}" start $command $args &
 362     wait_until_done $!
 363   ;;
 364
 365 (*)
 366   echo $usage
 367   exit 1
 368   ;;
 369 esac