354 lines
12 KiB
Bash
354 lines
12 KiB
Bash
#!/bin/sh
|
|
#
|
|
#
|
|
# OpenStack Cinder Volume (cinder-volume)
|
|
#
|
|
# Description: Manages an OpenStack Volumes (cinder-volume) process as an HA resource
|
|
#
|
|
# Authors: Sébastien Han
|
|
# Mainly inspired by the Glance API resource agent written by Martin Gerhard Loschwitz from Hastexo: http://goo.gl/whLpr
|
|
#
|
|
# Support: openstack@lists.launchpad.net
|
|
# License: Apache Software License (ASL) 2.0
|
|
#
|
|
#
|
|
# See usage() function below for more details ...
|
|
#
|
|
# OCF instance parameters:
|
|
# OCF_RESKEY_binary
|
|
# OCF_RESKEY_config
|
|
# OCF_RESKEY_user
|
|
# OCF_RESKEY_pid
|
|
# OCF_RESKEY_amqp_server_port
|
|
# OCF_RESKEY_additional_parameters
|
|
#######################################################################
|
|
# Initialization:
|
|
|
|
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
|
|
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
|
|
|
|
#######################################################################
|
|
|
|
# Fill in some defaults if no values are specified
|
|
|
|
OCF_RESKEY_binary_default="cinder-volume"
|
|
OCF_RESKEY_config_default="/etc/cinder/cinder.conf"
|
|
OCF_RESKEY_user_default="cinder"
|
|
OCF_RESKEY_pid_default="$HA_RSCTMP/$OCF_RESOURCE_INSTANCE.pid"
|
|
OCF_RESKEY_amqp_server_port_default="5672"
|
|
OCF_RESKEY_multibackend_default="false"
|
|
|
|
: ${OCF_RESKEY_binary=${OCF_RESKEY_binary_default}}
|
|
: ${OCF_RESKEY_config=${OCF_RESKEY_config_default}}
|
|
: ${OCF_RESKEY_user=${OCF_RESKEY_user_default}}
|
|
: ${OCF_RESKEY_pid=${OCF_RESKEY_pid_default}}
|
|
: ${OCF_RESKEY_amqp_server_port=${OCF_RESKEY_amqp_server_port_default}}
|
|
: ${OCF_RESKEY_multibackend=${OCF_RESKEY_multibackend_default}}
|
|
|
|
#######################################################################
|
|
|
|
usage() {
|
|
cat <<UEND
|
|
usage: $0 (start|stop|validate-all|meta-data|status|monitor)
|
|
|
|
$0 manages an OpenStack Cinder Volume (cinder-volume) process as an HA resource
|
|
|
|
The 'start' operation starts the volume service.
|
|
The 'stop' operation stops the volume service.
|
|
The 'validate-all' operation reports whether the parameters are valid
|
|
The 'meta-data' operation reports this RA's meta-data information
|
|
The 'status' operation reports whether the volume service is running
|
|
The 'monitor' operation reports whether the volume service seems to be working
|
|
|
|
UEND
|
|
}
|
|
|
|
meta_data() {
|
|
cat <<END
|
|
<?xml version="1.0"?>
|
|
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
|
|
<resource-agent name="cinder-volume">
|
|
<version>1.0</version>
|
|
|
|
<longdesc lang="en">
|
|
Resource agent for the OpenStack Cinder Volume (cinder-volume)
|
|
May manage a cinder-volume instance or a clone set that
|
|
creates a distributed cinder-volume cluster.
|
|
</longdesc>
|
|
<shortdesc lang="en">Manages the OpenStack Cinder Volume (cinder-volume)</shortdesc>
|
|
<parameters>
|
|
|
|
<parameter name="binary" unique="0" required="0">
|
|
<longdesc lang="en">
|
|
Location of the OpenStack Cinder Volume server binary (cinder-volume)
|
|
</longdesc>
|
|
<shortdesc lang="en">OpenStack Cinder Volume server binary (cinder-volume)</shortdesc>
|
|
<content type="string" default="${OCF_RESKEY_binary_default}" />
|
|
</parameter>
|
|
|
|
<parameter name="config" unique="0" required="0">
|
|
<longdesc lang="en">
|
|
Location of the OpenStack Cinder Volume (cinder-volume) configuration file
|
|
</longdesc>
|
|
<shortdesc lang="en">OpenStack Cinder Volume (cinder-volume) config file</shortdesc>
|
|
<content type="string" default="${OCF_RESKEY_config_default}" />
|
|
</parameter>
|
|
|
|
<parameter name="user" unique="0" required="0">
|
|
<longdesc lang="en">
|
|
User running OpenStack Cinder Volume (cinder-volume)
|
|
</longdesc>
|
|
<shortdesc lang="en">OpenStack Cinder Volume (cinder-volume) user</shortdesc>
|
|
<content type="string" default="${OCF_RESKEY_user_default}" />
|
|
</parameter>
|
|
|
|
<parameter name="pid" unique="0" required="0">
|
|
<longdesc lang="en">
|
|
The pid file to use for this OpenStack Cinder Volume (cinder-volume) instance
|
|
</longdesc>
|
|
<shortdesc lang="en">OpenStack Cinder Volume (cinder-volume) pid file</shortdesc>
|
|
<content type="string" default="${OCF_RESKEY_pid_default}" />
|
|
</parameter>
|
|
|
|
<parameter name="amqp_server_port" unique="0" required="0">
|
|
<longdesc lang="en">
|
|
The listening port number of the AMQP server. Mandatory to perform a monitor check
|
|
</longdesc>
|
|
<shortdesc lang="en">AMQP listening port</shortdesc>
|
|
<content type="integer" default="${OCF_RESKEY_amqp_server_port_default}" />
|
|
</parameter>
|
|
|
|
<parameter name="multibackend" unique="0" required="0">
|
|
<longdesc lang="en">
|
|
If the multi-backend is enabled, the monitor check is slightly different since cinder-volume spawns one thread for each backend.
|
|
</longdesc>
|
|
<shortdesc lang="en">Multi Backend usage</shortdesc>
|
|
<content type="boolean" default="${OCF_RESKEY_multibackend_default}" />
|
|
</parameter>
|
|
|
|
<parameter name="additional_parameters" unique="0" required="0">
|
|
<longdesc lang="en">
|
|
Additional parameters to pass on to the OpenStack Cinder Volume (cinder-volume)
|
|
</longdesc>
|
|
<shortdesc lang="en">Additional parameters for cinder-volume</shortdesc>
|
|
<content type="string" />
|
|
</parameter>
|
|
|
|
</parameters>
|
|
|
|
<actions>
|
|
<action name="start" timeout="10" />
|
|
<action name="stop" timeout="10" />
|
|
<action name="status" timeout="10" />
|
|
<action name="monitor" timeout="5" interval="10" />
|
|
<action name="validate-all" timeout="5" />
|
|
<action name="meta-data" timeout="5" />
|
|
</actions>
|
|
</resource-agent>
|
|
END
|
|
}
|
|
|
|
#######################################################################
|
|
# Functions invoked by resource manager actions
|
|
|
|
cinder_volume_validate() {
|
|
local rc
|
|
|
|
check_binary $OCF_RESKEY_binary
|
|
check_binary netstat
|
|
|
|
# A config file on shared storage that is not available
|
|
# during probes is OK.
|
|
if [ ! -f $OCF_RESKEY_config ]; then
|
|
if ! ocf_is_probe; then
|
|
ocf_log err "Config $OCF_RESKEY_config doesn't exist"
|
|
return $OCF_ERR_INSTALLED
|
|
fi
|
|
ocf_log_warn "Config $OCF_RESKEY_config not available during a probe"
|
|
fi
|
|
|
|
getent passwd $OCF_RESKEY_user >/dev/null 2>&1
|
|
rc=$?
|
|
if [ $rc -ne 0 ]; then
|
|
ocf_log err "User $OCF_RESKEY_user doesn't exist"
|
|
return $OCF_ERR_INSTALLED
|
|
fi
|
|
|
|
true
|
|
}
|
|
|
|
cinder_volume_status() {
|
|
local pid
|
|
local rc
|
|
|
|
if [ ! -f $OCF_RESKEY_pid ]; then
|
|
ocf_log info "OpenStack Cinder Volume (cinder-volume) is not running"
|
|
return $OCF_NOT_RUNNING
|
|
else
|
|
pid=`cat $OCF_RESKEY_pid`
|
|
fi
|
|
|
|
ocf_run -warn kill -s 0 $pid
|
|
rc=$?
|
|
if [ $rc -eq 0 ]; then
|
|
return $OCF_SUCCESS
|
|
else
|
|
ocf_log info "Old PID file found, but OpenStack Cinder Volume (cinder-volume) is not running"
|
|
return $OCF_NOT_RUNNING
|
|
fi
|
|
}
|
|
|
|
cinder_volume_monitor() {
|
|
local rc
|
|
local pid
|
|
local volume_amqp_check
|
|
|
|
cinder_volume_status
|
|
rc=$?
|
|
|
|
# If status returned anything but success, return that immediately
|
|
if [ $rc -ne $OCF_SUCCESS ]; then
|
|
return $rc
|
|
fi
|
|
|
|
# Grab cinder-volume PID
|
|
pid=`cat $OCF_RESKEY_pid`
|
|
|
|
if ocf_is_true "$OCF_RESKEY_multibackend"; then
|
|
# Grab the child's PIDs
|
|
for i in `ps -o pid --no-headers --ppid $pid`
|
|
do
|
|
volume_amqp_check=`netstat -punt | grep -s "$OCF_RESKEY_amqp_server_port" | grep -s "$i" | grep -qs "ESTABLISHED"`
|
|
rc=$?
|
|
if [ $rc -ne 0 ]; then
|
|
ocf_log err "This child process from Cinder Volume is not connected to the AMQP server: $rc"
|
|
return $OCF_NOT_RUNNING
|
|
fi
|
|
done
|
|
else
|
|
# Check the connections according to the PID
|
|
# We are sure to hit the scheduler process and not other nova process with the same connection behavior (for example nova-cert)
|
|
# check the connections according to the PID
|
|
|
|
# [Robert] This isn't working because it greps wrong pid. Need to look into it later. #
|
|
# volume_amqp_check=`netstat -punt | grep -s "$OCF_RESKEY_amqp_server_port" | grep -s "$pid" | grep -qs "ESTABLISHED"`
|
|
volume_amqp_check=`netstat -punt | grep -s "$OCF_RESKEY_amqp_server_port" | grep -qs "ESTABLISHED"`
|
|
rc=$?
|
|
if [ $rc -ne 0 ]; then
|
|
ocf_log err "Cinder Volume is not connected to the AMQP server: $rc"
|
|
return $OCF_NOT_RUNNING
|
|
fi
|
|
fi
|
|
|
|
ocf_log debug "OpenStack Cinder Volume (cinder-volume) monitor succeeded"
|
|
return $OCF_SUCCESS
|
|
}
|
|
|
|
cinder_volume_start() {
|
|
local rc
|
|
|
|
cinder_volume_status
|
|
rc=$?
|
|
if [ $rc -eq $OCF_SUCCESS ]; then
|
|
ocf_log info "OpenStack Cinder Volume (cinder-volume) already running"
|
|
return $OCF_SUCCESS
|
|
fi
|
|
|
|
# run the actual cinder-volume daemon. Don't use ocf_run as we're sending the tool's output
|
|
# straight to /dev/null anyway and using ocf_run would break stdout-redirection here.
|
|
su ${OCF_RESKEY_user} -s /bin/sh -c "${OCF_RESKEY_binary} --config-file=$OCF_RESKEY_config \
|
|
$OCF_RESKEY_additional_parameters"' >> /dev/null 2>&1 & echo $!' > $OCF_RESKEY_pid
|
|
|
|
# Spin waiting for the server to come up.
|
|
# Let the CRM/LRM time us out if required
|
|
while true; do
|
|
cinder_volume_monitor
|
|
rc=$?
|
|
[ $rc -eq $OCF_SUCCESS ] && break
|
|
if [ $rc -ne $OCF_NOT_RUNNING ]; then
|
|
ocf_log err "OpenStack Cinder Volume (cinder-volume) start failed"
|
|
exit $OCF_ERR_GENERIC
|
|
fi
|
|
sleep 1
|
|
done
|
|
|
|
ocf_log info "OpenStack Cinder Volume (cinder-volume) started"
|
|
return $OCF_SUCCESS
|
|
}
|
|
|
|
cinder_volume_stop() {
|
|
local rc
|
|
local pid
|
|
|
|
cinder_volume_status
|
|
rc=$?
|
|
if [ $rc -eq $OCF_NOT_RUNNING ]; then
|
|
ocf_log info "OpenStack Cinder Volume (cinder-volume) already stopped"
|
|
return $OCF_SUCCESS
|
|
fi
|
|
|
|
# Try SIGTERM
|
|
pid=`cat $OCF_RESKEY_pid`
|
|
ocf_run kill -s TERM $pid
|
|
rc=$?
|
|
if [ $rc -ne 0 ]; then
|
|
ocf_log err "OpenStack Cinder Volume (cinder-volume) couldn't be stopped"
|
|
exit $OCF_ERR_GENERIC
|
|
fi
|
|
|
|
# stop waiting
|
|
shutdown_timeout=15
|
|
if [ -n "$OCF_RESKEY_CRM_meta_timeout" ]; then
|
|
shutdown_timeout=$((($OCF_RESKEY_CRM_meta_timeout/1000)-5))
|
|
fi
|
|
count=0
|
|
while [ $count -lt $shutdown_timeout ]; do
|
|
cinder_volume_status
|
|
rc=$?
|
|
if [ $rc -eq $OCF_NOT_RUNNING ]; then
|
|
break
|
|
fi
|
|
count=`expr $count + 1`
|
|
sleep 1
|
|
ocf_log debug "OpenStack Cinder Volume (cinder-volume) still hasn't stopped yet. Waiting ..."
|
|
done
|
|
|
|
cinder_volume_status
|
|
rc=$?
|
|
if [ $rc -ne $OCF_NOT_RUNNING ]; then
|
|
# SIGTERM didn't help either, try SIGKILL
|
|
ocf_log info "OpenStack Cinder Volume (cinder-volume) failed to stop after ${shutdown_timeout}s \
|
|
using SIGTERM. Trying SIGKILL ..."
|
|
ocf_run kill -s KILL $pid
|
|
fi
|
|
|
|
ocf_log info "OpenStack Cinder Volume (cinder-volume) stopped"
|
|
|
|
rm -f $OCF_RESKEY_pid
|
|
|
|
return $OCF_SUCCESS
|
|
}
|
|
|
|
#######################################################################
|
|
|
|
case "$1" in
|
|
meta-data) meta_data
|
|
exit $OCF_SUCCESS;;
|
|
usage|help) usage
|
|
exit $OCF_SUCCESS;;
|
|
esac
|
|
|
|
# Anything except meta-data and help must pass validation
|
|
cinder_volume_validate || exit $?
|
|
|
|
# What kind of method was invoked?
|
|
case "$1" in
|
|
start) cinder_volume_start;;
|
|
stop) cinder_volume_stop;;
|
|
status) cinder_volume_status;;
|
|
monitor) cinder_volume_monitor;;
|
|
validate-all) ;;
|
|
*) usage
|
|
exit $OCF_ERR_UNIMPLEMENTED;;
|
|
esac
|