resource-agents 1:4.1.0~rc1-1ubuntu1 / usr / lib / ocf / resource.d / heartbeat / Xen

#!/bin/sh
#
#
# Support:      users@clusterlabs.org
# License:      GNU General Public License (GPL)
#
#	Resource Agent for the Xen Hypervisor.
#	Manages Xen virtual machine instances by
#	mapping cluster resource start and stop,
#	to Xen create and shutdown, respectively.
#
#	usage: $0  {start|stop|status|monitor|meta-data}
#
#	  OCF parameters are as below:
#		OCF_RESKEY_xmfile
#			Absolute path to the Xen control file,
#			for this virtual machine.
#		OCF_RESKEY_allow_mem_management
#			Change memory usage on start/stop/migration
#			of virtual machine
#		OCF_RESKEY_reserved_Dom0_memory
#			minimum memory reserved for domain 0
#		OCF_RESKEY_monitor_scripts
#			scripts to monitor services within the
#			virtual domain

#######################################################################
# Initialization:

: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs

#######################################################################

usage() {
	cat <<-END
	usage: $0 {start|stop|status|monitor|meta-data|validate-all}
END
}


: ${OCF_RESKEY_xmfile=/etc/xen/vm/MyDomU}
: ${OCF_RESKEY_shutdown_acpi=0}
: ${OCF_RESKEY_allow_mem_management=0}
: ${OCF_RESKEY_reserved_Dom0_memory=512}

# use /etc/default/xen
xentool=$(/usr/lib/xen-common/bin/xen-toolstack)

meta_data() {
	cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="Xen">
<version>1.0</version>

<longdesc lang="en">
Resource Agent for the Xen Hypervisor.
Manages Xen virtual machine instances by mapping cluster resource
start and stop, to Xen create and shutdown, respectively.

A note on names

We will try to extract the name from the config file (the xmfile
attribute). If you use a simple assignment statement, then you
should be fine. Otherwise, if there's some python acrobacy
involved such as dynamically assigning names depending on other
variables, and we will try to detect this, then please set the
name attribute. You should also do that if there is any chance of
a pathological situation where a config file might be missing,
for example if it resides on a shared storage. If all fails, we
finally fall back to the instance id to preserve backward
compatibility.

Para-virtualized guests can also be migrated by enabling the
meta_attribute allow-migrate.

</longdesc>
<shortdesc lang="en">Manages Xen unprivileged domains (DomUs)</shortdesc>

<parameters>

<parameter name="xmfile" unique="0" required="1">
<longdesc lang="en">
Absolute path to the Xen control file,
for this virtual machine.
</longdesc>
<shortdesc lang="en">Xen control file</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="name" unique="0" required="0">
<longdesc lang="en">
Name of the virtual machine.
</longdesc>
<shortdesc lang="en">Xen DomU name</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="shutdown_timeout">
<longdesc lang="en">
The Xen agent will first try an orderly shutdown using xl shutdown.
Should this not succeed within this timeout, the agent will escalate to
xl destroy, forcibly killing the node.

If this is not set, it will default to two-third of the stop action
timeout.

Setting this value to 0 forces an immediate destroy.

</longdesc>
<shortdesc lang="en">Shutdown escalation timeout</shortdesc>
<content type="string" default="" />
</parameter>
<parameter name="shutdown_acpi" unique="0" required="0">
<longdesc lang="en">
Handle shutdown by simulating an ACPI power button event.
Enable this to allow graceful shutdown for HVM domains
without installed PV drivers.
</longdesc>
<shortdesc lang="en">Simulate power button event on shutdown</shortdesc>
<content type="boolean" default="0" />
</parameter>
<parameter name="allow_mem_management" unique="0" required="0">
<longdesc lang="en">
This parameter enables dynamic adjustment of memory for start
and stop actions used for Dom0 and the DomUs. The default is
to not adjust memory dynamically.
</longdesc>
<shortdesc lang="en">Use dynamic memory management</shortdesc>
<content type="boolean" default="0" />
</parameter>

<parameter name="node_ip_attribute">
<longdesc lang="en">
In case of a live migration, the system will default to using the IP
address associated with the hostname via DNS or /etc/hosts.

This parameter allows you to specify a node attribute that will be
queried instead for the target node, overriding the IP address. This
allows you to use a dedicated network for live migration traffic to a
specific node.

Warning: make very sure the IP address does point to the right node. Or
else the live migration will end up somewhere else, greatly confusing
the cluster and causing havoc.
</longdesc>
<shortdesc lang="en">Node attribute containing target IP address</shortdesc>
<content type="string" default="" />
</parameter>

<parameter name="reserved_Dom0_memory" unique="0" required="0">
<longdesc lang="en">
In case memory management is used, this parameter
defines the minimum amount of memory to be reserved
for the dom0. The default minimum memory is 512MB.
</longdesc>
<shortdesc lang="en">Minimum Dom0 memory</shortdesc>
<content type="string" default="512" />
</parameter>
<parameter name="monitor_scripts" unique="0" required="0">
<longdesc lang="en">
To additionally monitor services within the unprivileged domain,
add this parameter with a list of scripts to monitor.
</longdesc>
<shortdesc lang="en">list of space separated monitor scripts</shortdesc>
<content type="string" default="" />
</parameter>

</parameters>

<actions>
<action name="start" timeout="60" />
<action name="stop" timeout="40" />
<action name="migrate_from" timeout="120" />
<action name="migrate_to" timeout="120" />
<action name="monitor" depth="0" timeout="30" interval="10" />
<action name="meta-data" timeout="5" />
<action name="validate-all" timeout="5" />
</actions>
</resource-agent>
END
}

Xen_Status() {
	if expr "x$xentool" : "x.*xl" >/dev/null; then
		$xentool list $1 >/dev/null 2>&1
		if [ $? -ne 0 ]; then
			return $OCF_NOT_RUNNING
		else
			return $OCF_SUCCESS
		fi
	fi
	if have_binary xen-list; then
		xen-list $1 2>/dev/null | grep -qs "State.*[-r][-b][-p]--" 2>/dev/null
		if [ $? -ne 0 ]; then
			return $OCF_NOT_RUNNING
		else
			return $OCF_SUCCESS
		fi
	fi
	STATUS=`$xentool list --long $1 2>/dev/null | grep status 2>/dev/null`
	if [ "X${STATUS}" != "X" ]; then
		# we have Xen 3.0.4 or higher
		STATUS_NOSPACES=`echo "$STATUS" | awk '{ print $1,$2}'`
		if [ "$STATUS_NOSPACES" = "(status 2)" -o "$STATUS_NOSPACES" = "(status 1)" ]; then
			return $OCF_SUCCESS
		else
			return $OCF_NOT_RUNNING
		fi
	else
		# we have Xen 3.0.3 or lower
		STATUS=`$xentool list --long $1 2>/dev/null | grep state 2>/dev/null`
		echo "${STATUS}" | grep -qs "[-r][-b][-p]---"
		if [ $? -ne 0 ]; then
			return $OCF_NOT_RUNNING
		else
			return $OCF_SUCCESS
		fi
	fi
}

# If the guest is rebooting, it may completely disappear from the
# list of defined guests, thus xl/xen-list would return with not
# running; apparently, this period lasts only for a second or
# two
# If a status returns not running, then test status
# again for 5 times (perhaps it'll show up)
Xen_Status_with_Retry() {
	local rc cnt=5

	Xen_Status $1
	rc=$?
	while [ $rc -eq $OCF_NOT_RUNNING -a $cnt -gt 0 ]; do
		case "$__OCF_ACTION" in
			stop)
				ocf_log debug "domain $1 reported as not running, waiting $cnt seconds ..."
				;;
			monitor)
				ocf_log warn "domain $1 reported as not running, but it is expected to be running! Retrying for $cnt seconds ..."
				;;
			*) : not reachable
				;;
		esac
		sleep 1
		Xen_Status $1
		rc=$?
		cnt=$((cnt-1))
	done
	return $rc
}

Xen_Adjust_Memory() {
	if ocf_is_true "${OCF_RESKEY_allow_mem_management}"; then
		CNTNEW=$1
		RUNNING=`Xen_List_running`
		RUNCNT=`Xen_Count_running`
		MAXMEM=`Xen_Total_Memory`
		if [ ${RUNCNT} -eq 0 -a ${CNTNEW} -eq 0 ]; then
			RUNCNT=1
		fi
		#NEWMEM=`echo "(${MAXMEM}-${OCF_RESKEY_reserved_Dom0_memory})/(${RUNCNT}+${CNTNEW})"|bc`
		NEWMEM=$(( (${MAXMEM} - ${OCF_RESKEY_reserved_Dom0_memory}) / (${RUNCNT} + ${CNTNEW} ) ))
		# do not rely on ballooning add dom0_mem=512 instead to force memory for dom0
		#$xentool mem-set Domain-0 ${OCF_RESKEY_reserved_Dom0_memory}
		for DOM in ${RUNNING}; do
			$xentool mem-set ${DOM} ${NEWMEM}
		done
		ocf_log info "Adjusted memory to: $NEWMEM, for the following $RUNCNT domains: $RUNNING"
	fi
}

Xen_List_all() {
	$xentool list | grep -v -e "Name" -e "Domain-0" | awk '{print $1}'
}
Xen_List_running() {
	ALL_DOMS=`Xen_List_all`
	for DOM in ${ALL_DOMS}; do
		if Xen_Status $DOM; then
			echo "${DOM} "
		fi
	done
}
Xen_Count_running() {
	Xen_List_running | wc -w
}

Xen_Monitor() {
	if ocf_is_probe; then
		Xen_Status ${DOMAIN_NAME}
	else
		Xen_Status_with_Retry ${DOMAIN_NAME}
	fi
	if [ $? -eq ${OCF_NOT_RUNNING} ]; then
		ocf_is_probe ||
		ocf_log err "Xen domain $DOMAIN_NAME stopped"
		return ${OCF_NOT_RUNNING}
	fi
	if [ "X${OCF_RESKEY_monitor_scripts}" = "X" ]; then
		return ${OCF_SUCCESS}
	fi
	for SCRIPT in ${OCF_RESKEY_monitor_scripts}; do
		$SCRIPT
		if [ $? -ne 0 ]; then
			return ${OCF_ERR_GENERIC}
		fi
	done
	return ${OCF_SUCCESS}
}

Xen_Total_Memory() {
	$xentool info | grep "^total_memory" | awk '{print $3}'
}

Xen_Start() {
	if Xen_Status ${DOMAIN_NAME}; then
		ocf_log info "Xen domain $DOMAIN_NAME already running."
		return $OCF_SUCCESS
	fi

	if [ ! -f "${OCF_RESKEY_xmfile}" ]; then
		ocf_log err "Config file ${OCF_RESKEY_xmfile} for $DOMAIN_NAME does not exist."
		return $OCF_ERR_INSTALLED
	fi

	if ocf_is_true "${OCF_RESKEY_allow_mem_management}"; then
		Xen_Adjust_Memory 1
		ocf_log info "New memory for virtual domains: ${NEWMEM}"
		sed -i -e "/^memory=/ s/^memory=.*/memory=${NEWMEM}/" ${OCF_RESKEY_xmfile}
		$xentool mem-set ${DOMAIN_NAME} ${NEWMEM}
	fi

	# the latest xl management tool is squeamish about some
	# characters in a name (the vm name is xen-f):
	# /etc/xen/vm/xen-f:15: config parsing error near `xen':
	# syntax error, unexpected IDENT, expecting STRING or NUMBER
	# or '['
	# /etc/xen/vm/xen-f:15: config parsing error near `-f': lexical error
	#
	# the older xm management tool cannot digest quotes (see
	# https://developerbugs.linuxfoundation.org/show_bug.cgi?id=2671)
	#
	# hence the following
	if expr "x$xentool" : "x.*xl" >/dev/null; then
		$xentool create ${OCF_RESKEY_xmfile} name=\"$DOMAIN_NAME\"
	else
		$xentool create ${OCF_RESKEY_xmfile} name="$DOMAIN_NAME"
	fi
	rc=$?

	if [ $rc -ne 0 ]; then
		return $OCF_ERR_GENERIC
	else
		if ocf_is_true "${OCF_RESKEY_allow_mem_management}"; then
			$xentool mem-set ${DOMAIN_NAME} ${NEWMEM}
		fi
	fi
	while sleep 1; do
		Xen_Monitor && return $OCF_SUCCESS
	done
}

xen_domain_stop() {
	local dom=$1
	local timeout

	if [ -n "$OCF_RESKEY_shutdown_timeout" ]; then
		timeout=$OCF_RESKEY_shutdown_timeout
	elif [ -n "$OCF_RESKEY_CRM_meta_timeout" ]; then
		# Allow 2/3 of the action timeout for the orderly shutdown
		# (The origin unit is ms, hence the conversion)
		timeout=$((OCF_RESKEY_CRM_meta_timeout/1500))
	else
		timeout=60
	fi

	if [ "$timeout" -gt 0 ]; then
		ocf_log info "Xen domain $dom will be stopped (timeout: ${timeout}s)"
		if ocf_is_true "${OCF_RESKEY_shutdown_acpi}"; then
			$xentool trigger $dom power
		else
			$xentool shutdown $dom
		fi

		while Xen_Status $dom && [ "$timeout" -gt 0 ]; do
			ocf_log debug "$dom still not stopped. Waiting..."
			timeout=$((timeout-1))
			sleep 1
		done
	fi

	if [ "$timeout" -eq 0 ]; then
		while Xen_Status $dom; do
			ocf_log warn "Xen domain $dom will be destroyed!"
			$xenkill $dom
			sleep 1
		done
		# Note: This does not give up. stop isn't allowed to to fail.
		# If $xentool destroy fails, stop will eventually timeout.
		# This is the correct behaviour.
	fi

	ocf_log info "Xen domain $dom stopped."
}

Xen_Stop() {
	local vm
	if Xen_Status_with_Retry ${DOMAIN_NAME}; then
		vm=${DOMAIN_NAME}
	elif Xen_Status migrating-${DOMAIN_NAME}; then
		ocf_log info "Xen domain $DOMAIN_NAME is migrating"
		vm="migrating-${DOMAIN_NAME}"
	else
		ocf_log info "Xen domain $DOMAIN_NAME already stopped."
	fi

	if [ "$vm" ]; then
		xen_domain_stop $vm
	else
		# It is supposed to be gone, but there have been situations where
		# $xentool list / xen-list showed it as stopped but it was still
		# instantiated. Nuke it once more to make sure:
		$xenkill ${DOMAIN_NAME}
	fi

	Xen_Adjust_Memory 0
	return $OCF_SUCCESS
}

Xen_Migrate_To() {
	target_node="$OCF_RESKEY_CRM_meta_migrate_target"
	target_attr="$OCF_RESKEY_node_ip_attribute"
	target_addr="$target_node"

	if Xen_Status ${DOMAIN_NAME}; then
		ocf_log info "$DOMAIN_NAME: Starting $xentool migrate to $target_node"

		if [ -n "$target_attr" ]; then
			nodevalue=`crm_attribute --type nodes --node-uname $target_node --attr-name $target_attr --get-value -q`
			if [ -n "${nodevalue}" -a "${nodevalue}" != "(null)" ]; then
				target_addr="$nodevalue"
				ocf_log info "$DOMAIN_NAME: $target_node is using address $target_addr"
			fi
		fi

		if expr "x$xentool" : "x.*xm" >/dev/null; then
			$xentool migrate --live $DOMAIN_NAME $target_addr
		else
			$xentool migrate $DOMAIN_NAME $target_addr
		fi
		rc=$?
		if [ $rc -ne 0 ]; then
			ocf_log err "$DOMAIN_NAME: $xentool migrate to $target_node failed: $rc"
			return $OCF_ERR_GENERIC
		else
			Xen_Adjust_Memory 0
			ocf_log info "$DOMAIN_NAME: $xentool migrate to $target_node succeeded."
			return $OCF_SUCCESS
		fi
	else
		ocf_log err "$DOMAIN_NAME: migrate_to: Not active locally!"
		return $OCF_ERR_GENERIC
	fi
}

Xen_Migrate_From() {
	if [ -n "$OCF_RESKEY_CRM_meta_timeout" ]; then
		# Allow 2/3 of the action timeout for status to stabilize
		# (The origin unit is ms, hence the conversion)
		timeout=$((OCF_RESKEY_CRM_meta_timeout/1500))
	else
		timeout=10		# should be plenty
	fi

	while ! Xen_Status ${DOMAIN_NAME} && [ $timeout -gt 0 ]; do
		ocf_log debug "$DOMAIN_NAME: Not yet active locally, waiting (timeout: ${timeout}s)"
		timeout=$((timeout-1))
		sleep 1
	done

	if Xen_Status ${DOMAIN_NAME}; then
		Xen_Adjust_Memory 0
		ocf_log info "$DOMAIN_NAME: Active locally, migration successful"
		return $OCF_SUCCESS
	else
		ocf_log err "$DOMAIN_NAME: Not active locally, migration failed!"
		return $OCF_ERR_GENERIC
	fi
}

Xen_Validate_All() {
	return $OCF_SUCCESS
}

if [ $# -ne 1 ]; then
	usage
	exit $OCF_ERR_ARGS
fi

case $1 in
	meta-data)
		meta_data
		exit $OCF_SUCCESS
		;;
	usage)
		usage
		exit $OCF_SUCCESS
		;;
esac

# the name business:
#
# 1. use the name attribute, or
# 2. find the name in the config file (if it exists) and use that
# unless it contains funny characters such as '%' or space, or
# 3. use the OCF_RESOURCE_INSTANCE

if [ x"${OCF_RESKEY_name}" != x ]; then
	DOMAIN_NAME="${OCF_RESKEY_name}"
else
	if [ -f "${OCF_RESKEY_xmfile}" ]; then
		DOMAIN_NAME=`awk '$1~/^name(=|$)/{print}' ${OCF_RESKEY_xmfile} | sed 's/.*=[[:space:]]*//' | tr -d "[\"']"`
		if echo "$DOMAIN_NAME" | grep -qs '[%[:space:]]'; then
			DOMAIN_NAME=""
		fi
	fi
	DOMAIN_NAME=${DOMAIN_NAME:-${OCF_RESOURCE_INSTANCE}}
fi

for binary in sed awk; do
	check_binary $binary
done

if have_binary xen-destroy ; then
	xenkill="xen-destroy"
else
	xenkill="$xentool destroy"
fi

if [ -n "$OCF_RESKEY_shutdown_timeout" ]; then
	ocf_is_decimal "$OCF_RESKEY_shutdown_timeout" || {
		ocf_log err "shutdown_timeout must be a number"
		exit $OCF_ERR_CONFIGURED
	}
fi

case $1 in
	start)
		Xen_Start
		;;
	stop)
		Xen_Stop
		;;
	migrate_to)
		Xen_Migrate_To
		;;
	migrate_from)
		Xen_Migrate_From
		;;
	monitor)
		Xen_Monitor
		;;
	status)
		Xen_Status ${DOMAIN_NAME}
		;;
	validate-all)
		Xen_Validate_All
		;;
	*)
		usage
		exit $OCF_ERR_UNIMPLEMENTED
		;;
esac
exit $?