diff --git a/lrmd/pacemaker_remote.in b/lrmd/pacemaker_remote.in index 6f79fae78c..a356510c1a 100644 --- a/lrmd/pacemaker_remote.in +++ b/lrmd/pacemaker_remote.in @@ -1,157 +1,175 @@ #!/bin/bash # Authors: # Andrew Beekhof # # License: Revised BSD # chkconfig: - 99 01 # description: Pacemaker Cluster Manager # processname: pacemaker_remoted # ### BEGIN INIT INFO # Provides: pacemaker_remote # Required-Start: $network $remote_fs # Should-Start: $syslog # Required-Stop: $network $remote_fs # Default-Start: # Default-Stop: # Short-Description: Starts and stops the Pacemaker remote agent for non-cluster nodes # Description: Starts and stops the Pacemaker remote agent for non-cluster nodes ### END INIT INFO desc="Pacemaker Remote Agent" prog="pacemaker_remoted" cman=0 # set secure PATH PATH="/sbin:/bin:/usr/sbin:/usr/bin:@sbindir@" checkrc() { if [ $? = 0 ]; then success else failure fi } success() { echo -ne "[ OK ]\r" } failure() { echo -ne "[FAILED]\r" } status() { pid=$(pidof $1 2>/dev/null) - rtrn=$? + local rtrn=$? if [ $rtrn -ne 0 ]; then echo "$1 is stopped" + if [ -f "@localstatedir@/run/$prog.pid" ]; then + rtrn=1 + else + rtrn=3 + fi else echo "$1 (pid $pid) is running..." fi return $rtrn } if [ -d @CONFIGDIR@ ]; then [ -f @INITDIR@/functions ] && . @INITDIR@/functions set -a [ -f @CONFIGDIR@/pacemaker ] && . @CONFIGDIR@/pacemaker + [ -f @CONFIGDIR@/sbd ] && . @CONFIGDIR@/sbd set +a fi LOCK_DIR="." if [ -d "@localstatedir@/lock/subsys" ]; then LOCK_DIR="@localstatedir@/lock/subsys" elif [ -d "@localstatedir@/lock" ]; then LOCK_DIR="@localstatedir@/lock" fi [ -z "$LOCK_FILE" ] && LOCK_FILE="$LOCK_DIR/pacemaker_remote" +# Check if there is a valid watchdog-device configured in sbd config +if [ x != "x$SBD_WATCHDOG_DEV" -a "/dev/null" != "$SBD_WATCHDOG_DEV" -a -c "$SBD_WATCHDOG_DEV" ]; then + # enhance for unavailable chkconfig - don't touch sbd for now + if chkconfig --list sbd_remote_helper 2>/dev/null | grep -q ":on"; then + SBD_SERVICE=sbd_remote_helper + fi +fi + start() { echo -n "Starting $desc: " # most recent distributions use tmpfs for $@localstatedir@/run # to avoid to clean it up on every boot. # they also assume that init scripts will create # required subdirectories for proper operations mkdir -p "@localstatedir@/run" if status $prog > /dev/null 2>&1; then success else $prog > /dev/null 2>&1 & # Time to connect to corosync and fail sleep 5 if status $prog > /dev/null 2>&1; then touch "$LOCK_FILE" pidof $prog > "@localstatedir@/run/$prog.pid" success else failure rtrn=1 fi fi echo + + [ "x$SBD_SERVICE" != x ] && service $SBD_SERVICE start } stop() { if status $prog > /dev/null 2>&1; then echo -n "Signaling $desc to terminate: " kill -TERM $(pidof $prog) > /dev/null 2>&1 success echo echo -n "Waiting for $desc to unload:" while status $prog > /dev/null 2>&1; do sleep 1 echo -n "." done fi rm -f "$LOCK_FILE" rm -f "@localstatedir@/run/$prog.pid" success echo + + [ "x$SBD_SERVICE" != x ] && service $SBD_SERVICE stop } rtrn=0 case "$1" in start) start ;; restart|reload|force-reload) stop start ;; condrestart|try-restart) if status $prog > /dev/null 2>&1; then stop start rtrn=$? fi ;; status) status $prog rtrn=$? ;; stop) stop rtrn=$? ;; *) echo "usage: $0 {start|stop|restart|reload|force-reload|condrestart|try-restart|status}" rtrn=2 ;; esac exit $rtrn diff --git a/mcp/pacemaker.in b/mcp/pacemaker.in index 6e90645a74..0ff1ac137f 100644 --- a/mcp/pacemaker.in +++ b/mcp/pacemaker.in @@ -1,285 +1,300 @@ #!/bin/bash # Authors: # Andrew Beekhof # Fabio M. Di Nitto # # License: Revised BSD # chkconfig: - 99 01 # description: Pacemaker Cluster Manager # processname: pacemakerd # ### BEGIN INIT INFO # Provides: pacemaker # Required-Start: $network $remote_fs corosync # Should-Start: $syslog # Required-Stop: $network $remote_fs corosync # Default-Start: # Default-Stop: # Short-Description: Starts and stops Pacemaker Cluster Manager. # Description: Starts and stops Pacemaker Cluster Manager. ### END INIT INFO desc="Pacemaker Cluster Manager" prog="pacemakerd" # set secure PATH PATH="/sbin:/bin:/usr/sbin:/usr/bin:@sbindir@" checkrc() { if [ $? = 0 ]; then success else failure fi } success() { echo -ne "[ OK ]\r" } failure() { echo -ne "[FAILED]\r" } log() { logger -t pacemaker -p daemon.notice "$*" } notify() { log "$*" echo -n "$*" } status() { pid=$(pidof $1 2>/dev/null) local rtrn=$? if [ $rtrn -ne 0 ]; then echo "$1 is stopped" + if [ -f "@localstatedir@/run/$prog.pid" ]; then + rtrn=1 + else + rtrn=3 + fi else echo "$1 (pid $pid) is running..." fi return $rtrn } if [ -d @CONFIGDIR@ ]; then [ -f @INITDIR@/functions ] && . @INITDIR@/functions set -a [ -f @CONFIGDIR@/pacemaker ] && . @CONFIGDIR@/pacemaker + [ -f @CONFIGDIR@/sbd ] && . @CONFIGDIR@/sbd set +a fi LOCK_DIR="." if [ -d "@localstatedir@/lock/subsys" ]; then LOCK_DIR="@localstatedir@/lock/subsys" elif [ -d "@localstatedir@/lock" ]; then LOCK_DIR="@localstatedir@/lock" fi [ -z "$LOCK_FILE" ] && LOCK_FILE="$LOCK_DIR/pacemaker" # Unless specified otherwise, assume cman is in use if cluster.conf exists if [ x = "x$PCMK_STACK" -a -f "@sysconfdir@/cluster/cluster.conf" ]; then PCMK_STACK=cman fi +# Check if there is a valid watchdog-device configured in sbd config +if [ x != "x$SBD_WATCHDOG_DEV" -a "/dev/null" != "$SBD_WATCHDOG_DEV" -a -c "$SBD_WATCHDOG_DEV" ]; then + # enhance for unavailable chkconfig - don't touch sbd for now + if chkconfig --list sbd_helper 2>/dev/null | grep -q ":on"; then + SBD_SERVICE=sbd_helper + fi +fi + start() { notify "Starting $desc" # most recent distributions use tmpfs for $@localstatedir@/run # to avoid to clean it up on every boot. # they also assume that init scripts will create # required subdirectories for proper operations mkdir -p "@localstatedir@/run" if status $prog > /dev/null 2>&1; then success else $prog > /dev/null 2>&1 & # Time to connect to corosync and fail sleep 5 if status $prog > /dev/null 2>&1; then touch "$LOCK_FILE" pidof $prog > "@localstatedir@/run/$prog.pid" success else failure rtrn=1 fi fi echo } cman_pre_start() { # start cman if it's not running service cman status >/dev/null 2>&1 if [ $? -ne 0 ]; then service cman start if [ $? -ne 0 ]; then notify "Aborting startup of $desc" echo exit 1 fi fi # start cman's friends if they're not running but were configured to start automatically - for cservice in cmirrord clvmd gfs2; do + for cservice in cmirrord clvmd gfs2 $SBD_SERVICE; do chkconfig --list $cservice 2>/dev/null | grep -q ":on" if [ $? -eq 0 ]; then service $cservice status >/dev/null 2>&1 if [ $? -ne 0 ]; then log "Attempting to start $cservice" service $cservice start fi fi done } cman_pre_stop() { # if cman is not running there is nothing we need to do here service cman status >/dev/null 2>&1 if [ $? -ne 0 ]; then return fi has_lvm=`crm_resource -c | grep Resource: | grep LVM` # migrate resources to another node or shut them down cname=`crm_node --name` crm_attribute -N $cname -n standby -v true -l reboot notify "Waiting for shutdown of managed resources" while [ 1 = 1 ]; do # 0x0000000000000002 means managed active=`crm_resource -c | grep Resource: | grep 0x...............[2367] | awk '{print $9}' | grep "^${cname}$" | wc -l` if [ $active = 0 ]; then break; fi sleep 2 echo -n "." done success echo if [ -d /sys/kernel/dlm/ ]; then lockspace="$(ls -1 /sys/kernel/dlm/)" if [ -n "$lockspace" ]; then notify "DLM lockspace still in use" echo "" for cservice in gfs2 clvmd cmirrord; do service $cservice status >/dev/null 2>&1 if [ $? -eq 0 ]; then if [ -n "$has_lvm" ] && [ "$cservice" = "clvmd" ]; then # allow HA-LVM to take a lock on vg/lv before clvmd can exit notify "Waiting for LVM services to start somewhere else" sleep 15 success fi log "Attempting to shutdown $cservice" service $cservice stop fi done fi fi notify "Leaving fence domain" fence_tool leave -w 10 checkrc fenced=$(pidof fenced) notify "Stopping fenced $fenced" kill -KILL $fenced > /dev/null 2>&1 checkrc } stop() { shutdown_prog=$prog if ! status $prog > /dev/null 2>&1; then shutdown_prog="crmd" fi if status $shutdown_prog > /dev/null 2>&1; then notify "Signaling $desc to terminate" kill -TERM $(pidof $prog) > /dev/null 2>&1 checkrc echo notify "Waiting for cluster services to unload" while status $prog > /dev/null 2>&1; do sleep 1 echo -n "." done else echo -n "$desc is already stopped" fi rm -f "$LOCK_FILE" rm -f "@localstatedir@/run/$prog.pid" killall -q -9 'crmd stonithd attrd cib lrmd pacemakerd' success echo } rtrn=0 case "$1" in start) # For consistency with stop [ "$PCMK_STACK" = cman ] && cman_pre_start start ;; restart|reload|force-reload) stop start ;; condrestart|try-restart) if status $prog > /dev/null 2>&1; then stop start fi ;; status) status $prog rtrn=$? ;; stop) # # stonithd needs to be around until fenced is stopped # fenced can't be stopped until any resource using dlm is active # # So: # 1. put the node into standby # 2. wait for all resources to be stopped # 3. stop fenced and anything that needs it (borrowed from the cman script) # 4. stop pacemaker # 5. stop the rest of cman (so it doesn't end up half up/down) # [ "$PCMK_STACK" = cman ] && cman_pre_stop stop + [ "x$SBD_SERVICE" != x ] && service $SBD_SERVICE stop # Stop cman if needed, unless --skip-cman is specified (which allows # higher-level tooling to stop pacemaker on all nodes, then stop cman # on all nodes, to maintain quorum for DLM-based resources while # pacemaker shuts down). [ "$PCMK_STACK" = cman ] && [ "$2" != "--skip-cman" ] && service cman stop ;; *) echo "usage: $0 {start|stop|restart|reload|force-reload|condrestart|try-restart|status}" rtrn=2 ;; esac exit $rtrn