openstack-resource-agents/ocf/nova-network

397 lines
12 KiB
Bash

#!/bin/sh
#
#
# OpenStack Network Service (nova-network) (replaced by Neutron in Havana)
#
# Description: Manages an OpenStack Network Service (nova-network) process as
# an HA resource
#
# Authors: Sebastien Han & Emilien Macchi
# Mainly inspired by the Nova Scheduler resource agent written by Sebastien
# Han: http://goo.gl/s8hOU Which are also inspired by the resource agents
# written by Martin Gerhard Loschwitz from Hastexo: http://goo.gl/whLpr
#
# Support: openstack@lists.openstack.org
# 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_database_server_port
# OCF_RESKEY_amqp_server_port
# OCF_RESKEY_zeromq
# OCF_RESKEY_additional_parameters
#######################################################################
# Initialization:
: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
#######################################################################
# Fill in some defaults if no values are specified
OCF_RESKEY_binary_default="nova-network"
OCF_RESKEY_config_default="/etc/nova/nova.conf"
OCF_RESKEY_user_default="nova"
OCF_RESKEY_pid_default="$HA_RSCTMP/$OCF_RESOURCE_INSTANCE.pid"
OCF_RESKEY_database_server_port_default="3306"
OCF_RESKEY_amqp_server_port_default="5672"
OCF_RESKEY_zeromq_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}}
: ${OCF_RESKEY_database_server_port=${OCF_RESKEY_database_server_port_default}}
: ${OCF_RESKEY_amqp_server_port=${OCF_RESKEY_amqp_server_port_default}}
: ${OCF_RESKEY_zeromq=${OCF_RESKEY_zeromq_default}}
#######################################################################
usage() {
cat <<UEND
usage: $0 (start|stop|validate-all|meta-data|status|monitor)
$0 manages an OpenStack Network Service (nova-network) process as an HA resource
The 'start' operation starts the networking service.
The 'stop' operation stops the networking service.
The 'validate-all' operation reports whether the parameters are valid
The 'meta-data' operation reports this RA's meta-data information
The 'status' operation reports whether the networking service is running
The 'monitor' operation reports whether the networking service seems to be working
UEND
}
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="nova-network">
<version>1.0</version>
<longdesc lang="en">
Resource agent for the OpenStack Nova Network Service (nova-network)
May manage a nova-network instance or a clone set that
creates a distributed nova-network cluster.
</longdesc>
<shortdesc lang="en">Manages the OpenStack Network Service (nova-network)</shortdesc>
<parameters>
<parameter name="binary" unique="0" required="0">
<longdesc lang="en">
Location of the OpenStack Nova Network server binary (nova-network)
</longdesc>
<shortdesc lang="en">OpenStack Nova Network server binary (nova-network)</shortdesc>
<content type="string" default="${OCF_RESKEY_binary_default}" />
</parameter>
<parameter name="config" unique="0" required="0">
<longdesc lang="en">
Location of the OpenStack Network Service (nova-network) configuration file
</longdesc>
<shortdesc lang="en">OpenStack Nova Network (nova-network) config file</shortdesc>
<content type="string" default="${OCF_RESKEY_config_default}" />
</parameter>
<parameter name="user" unique="0" required="0">
<longdesc lang="en">
User running OpenStack Network Service (nova-network)
</longdesc>
<shortdesc lang="en">OpenStack Network Service (nova-network) user</shortdesc>
<content type="string" default="${OCF_RESKEY_user_default}" />
</parameter>
<parameter name="pid" unique="0" required="0">
<longdesc lang="en">
The pid file to use for this OpenStack Network Service (nova-network) instance
</longdesc>
<shortdesc lang="en">OpenStack Network Service (nova-network) pid file</shortdesc>
<content type="string" default="${OCF_RESKEY_pid_default}" />
</parameter>
<parameter name="database_server_port" unique="0" required="0">
<longdesc lang="en">
The listening port number of the database server. Mandatory to perform a monitor check
</longdesc>
<shortdesc lang="en">Database listening port</shortdesc>
<content type="integer" default="${OCF_RESKEY_database_server_port_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="${OCF_RESKEY_amqp_server_port_default}" />
</parameter>
<parameter name="zeromq" unique="0" required="0">
<longdesc lang="en">
If zeromq is used, this will disable the connection test to the AMQP server
</longdesc>
<shortdesc lang="en">Zero-MQ usage</shortdesc>
<content type="boolean" default="${OCF_RESKEY_zeromq_default}" />
</parameter>
<parameter name="additional_parameters" unique="0" required="0">
<longdesc lang="en">
Additional parameters to pass on to the OpenStack Network Service (nova-network)
</longdesc>
<shortdesc lang="en">Additional parameters for nova-network</shortdesc>
<content type="string" />
</parameter>
</parameters>
<actions>
<action name="start" timeout="20" />
<action name="stop" timeout="20" />
<action name="status" timeout="20" />
<action name="monitor" timeout="30" interval="20" />
<action name="validate-all" timeout="5" />
<action name="meta-data" timeout="5" />
</actions>
</resource-agent>
END
}
#######################################################################
# Functions invoked by resource manager actions
nova_network_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
}
nova_network_status() {
local pid
local rc
if [ ! -f $OCF_RESKEY_pid ]; then
ocf_log info "OpenStack Nova Network (nova-network) 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 Nova Network" \
"(nova-network) is not running"
return $OCF_NOT_RUNNING
fi
}
nova_network_monitor() {
local rc
local pid
local rc_db
local rc_amqp
local network_db_check
local network_amqp_check
nova_network_status
rc=$?
# If status returned anything but success, return that immediately
if [ $rc -ne $OCF_SUCCESS ]; then
return $rc
fi
# Check the connections according to the PID.
# We are sure to hit the scheduler process and not other nova process with
# the same connection behavior (for example nova-cert)
if ocf_is_true "$OCF_RESKEY_zeromq"; then
pid=`cat $OCF_RESKEY_pid`
network_db_check=`netstat -punt | \
grep -s "$OCF_RESKEY_database_server_port" | grep -s "$pid" | \
grep -qs "ESTABLISHED"`
rc_db=$?
if [ $rc_db -ne 0 ]; then
ocf_log err "Nova Network is not connected to the database" \
"server: $rc_db"
return $OCF_NOT_RUNNING
fi
else
pid=`cat $OCF_RESKEY_pid`
# check the connections according to the PID
network_db_check=`netstat -punt | \
grep -s "$OCF_RESKEY_database_server_port" | grep -s "$pid" | \
grep -qs "ESTABLISHED"`
rc_db=$?
network_amqp_check=`netstat -punt | \
grep -s "$OCF_RESKEY_amqp_server_port" | grep -s "$pid" | \
grep -qs "ESTABLISHED"`
rc_amqp=$?
if [ $rc_amqp -ne 0 ] || [ $rc_db -ne 0 ]; then
ocf_log err "Nova Network is not connected to the AMQP server" \
"and/or the database server: AMQP connection test returned" \
"$rc_amqp and database connection test returned $rc_db"
return $OCF_NOT_RUNNING
fi
fi
ocf_log debug "OpenStack Nova Network (nova-network) monitor succeeded"
return $OCF_SUCCESS
}
nova_network_start() {
local rc
nova_network_status
rc=$?
if [ $rc -eq $OCF_SUCCESS ]; then
ocf_log info "OpenStack Nova Network (nova-network) already running"
return $OCF_SUCCESS
fi
# run the actual nova-network 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 $!' \
> $OCF_RESKEY_pid
# Spin waiting for the server to come up.
# Let the CRM/LRM time us out if required
while true; do
nova_network_monitor
rc=$?
[ $rc -eq $OCF_SUCCESS ] && break
if [ $rc -ne $OCF_NOT_RUNNING ]; then
ocf_log err "OpenStack Nova Network (nova-network) start failed"
exit $OCF_ERR_GENERIC
fi
sleep 1
done
ocf_log info "OpenStack Nova Network (nova-network) started"
return $OCF_SUCCESS
}
nova_network_stop() {
local rc
local pid
nova_network_status
rc=$?
if [ $rc -eq $OCF_NOT_RUNNING ]; then
ocf_log info "OpenStack Nova Network (nova-network) already stopped"
return $OCF_SUCCESS
fi
# Try SIGTERM
pid=`cat $OCF_RESKEY_pid`
pid="$pid `ps ax | grep -v grep | grep 'dnsmasq' | cut -c1-5`"
ocf_run kill -s TERM $pid
rc=$?
if [ $rc -ne 0 ]; then
ocf_log err "OpenStack Nova Network (nova-network) 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 $shutdown_timeout ]; do
nova_network_status
rc=$?
if [ $rc -eq $OCF_NOT_RUNNING ]; then
break
fi
count=`expr $count + 1`
sleep 1
ocf_log debug "OpenStack Nova Network (nova-network) still hasn't" \
"stopped yet. Waiting ..."
done
nova_network_status
rc=$?
if [ $rc -ne $OCF_NOT_RUNNING ]; then
# SIGTERM didn't help either, try SIGKILL
ocf_log info "OpenStack Nova Network (nova-network) failed to stop" \
"after ${shutdown_timeout}s using SIGTERM. Trying SIGKILL ..."
ocf_run kill -s KILL $pid
fi
ocf_log info "OpenStack Nova Network (nova-network) 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
nova_network_validate || exit $?
# What kind of method was invoked?
case "$1" in
start)
nova_network_start
;;
stop)
nova_network_stop
;;
status)
nova_network_status
;;
monitor)
nova_network_monitor
;;
validate-all)
;;
*)
usage
exit $OCF_ERR_UNIMPLEMENTED
;;
esac