openstack-resource-agents/ocf/nova-scheduler

#!/bin/sh
#
#
# OpenStack Scheduler Service (nova-scheduler)
#
# Description:  Manages an OpenStack Scheduler Service (nova-scheduler) 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.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_monitor_binary
#   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-scheduler"
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 SchedulerService (nova-scheduler) process as an HA resource

        The 'start' operation starts the scheduler service.
        The 'stop' operation stops the scheduler 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 scheduler service is running
        The 'monitor' operation reports whether the scheduler 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-scheduler">
<version>1.0</version>

<longdesc lang="en">
Resource agent for the OpenStack Nova Scheduler Service (nova-scheduler)
May manage a nova-scheduler instance or a clone set that
creates a distributed nova-scheduler cluster.
</longdesc>
<shortdesc lang="en">Manages the OpenStack Scheduler Service (nova-scheduler)</shortdesc>
<parameters>

<parameter name="binary" unique="0" required="0">
<longdesc lang="en">
Location of the OpenStack Nova Scheduler server binary (nova-scheduler)
</longdesc>
<shortdesc lang="en">OpenStack Nova Scheduler server binary (nova-scheduler)</shortdesc>
<content type="string" default="${OCF_RESKEY_binary_default}" />
</parameter>

<parameter name="config" unique="0" required="0">
<longdesc lang="en">
Location of the OpenStack Scheduler Service (nova-scheduler) configuration file
</longdesc>
<shortdesc lang="en">OpenStack Nova Scheduler (nova-scheduler) 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 Scheduler Service (nova-scheduler)
</longdesc>
<shortdesc lang="en">OpenStack Scheduler Service (nova-scheduler) 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 Scheduler Service (nova-scheduler) instance
</longdesc>
<shortdesc lang="en">OpenStack Scheduler Service (nova-scheduler) 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. Use for monitoring purposes
</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. Use for monitoring purposes
</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. Use for monitoring purposes
</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 Scheduler Service (nova-scheduler)
</longdesc>
<shortdesc lang="en">Additional parameters for nova-scheduler</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_scheduler_check_port() {
# This function has been taken from the squid RA and improved a bit
# The length of the integer must be 4
# Examples of valid port: "1080", "0080"
# Examples of invalid port: "1080bad", "0", "0000", ""

    local int
    local cnt

    int="$1"
    cnt=${#int}
    echo $int |egrep -qx '[0-9]+(:[0-9]+)?(,[0-9]+(:[0-9]+)?)*'

    if [ $? -ne 0 ] || [ $cnt -ne 4 ]; then
        ocf_log err "Invalid port number: $1"
        exit $OCF_ERR_CONFIGURED
    fi
}

nova_scheduler_validate() {
    local rc

    check_binary $OCF_RESKEY_binary
    check_binary netstat
    nova_scheduler_check_port $OCF_RESKEY_database_server_port
    nova_scheduler_check_port $OCF_RESKEY_amqp_server_port

    # 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_scheduler_status() {
    local pid
    local rc

    if [ ! -f $OCF_RESKEY_pid ]; then
        ocf_log info "OpenStack Nova Scheduler (nova-scheduler) 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 Scheduler" \
            "(nova-scheduler) is not running"
        return $OCF_NOT_RUNNING
    fi
}

nova_scheduler_monitor() {
    local rc
    local pid
    local rc_db
    local rc_amqp
    local scheduler_db_check
    local scheduler_amqp_check

    nova_scheduler_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 $OCF_RESKEY_pid`
        scheduler_db_check=`netstat -punt | \
            grep -s "$OCF_RESKEY_database_server_port" | grep -s "$pid" | \
            grep -qs "ESTABLISHED"`
        rc_db=$?
        if [ $rc_db -ne 0 ]; then
            ocf_log err "Nova Scheduler is not connected to the database" \
                "server: $rc_db"
            return $OCF_NOT_RUNNING
        fi
    else
        pid=`cat $OCF_RESKEY_pid`
        scheduler_db_check=`netstat -punt | \
            grep -s "$OCF_RESKEY_database_server_port" | grep -s "$pid" | \
            grep -qs "ESTABLISHED"`
        rc_db=$?
        scheduler_amqp_check=`netstat -punt | \
            grep -s "$OCF_RESKEY_amqp_server_port" | grep -s "$pid" | \
            grep -qs "ESTABLISHED"`
        rc_amqp=$?
        if [ $rc_amqp -ne 0 ] || [ $rc_db -ne 0 ]; then
            ocf_log err "Nova Scheduler is not connected to the AMQP server" \
                "and/or the database server: AMQP connection test returned" \
                "$rc_amqp and database connection test returned $rc_db"
            return $OCF_NOT_RUNNING
        fi
    fi

    ocf_log debug "OpenStack Nova Scheduler (nova-scheduler) monitor succeeded"
    return $OCF_SUCCESS
}

nova_scheduler_start() {
    local rc

    nova_scheduler_status
    rc=$?
    if [ $rc -eq $OCF_SUCCESS ]; then
        ocf_log info "OpenStack Nova Scheduler (nova-scheduler) already" \
            "running"
        return $OCF_SUCCESS
    fi

    # run the actual nova-scheduler 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.
    while true; do
    nova_scheduler_monitor
    rc=$?
    [ $rc -eq $OCF_SUCCESS ] && break
    if [ $rc -ne $OCF_NOT_RUNNING ]; then
        ocf_log err "OpenStack Nova Scheduler (nova-scheduler) start failed"
        exit $OCF_ERR_GENERIC
    fi
    sleep 1
    done

    ocf_log info "OpenStack Nova Scheduler (nova-scheduler) started"
    return $OCF_SUCCESS
}

nova_scheduler_stop() {
    local rc
    local pid

    nova_scheduler_status
    rc=$?
    if [ $rc -eq $OCF_NOT_RUNNING ]; then
        ocf_log info "OpenStack Nova Scheduler (nova-scheduler) 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 Nova Scheduler (nova-scheduler) 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 $shutdown_timeout ]; do
        nova_scheduler_status
        rc=$?
        if [ $rc -eq $OCF_NOT_RUNNING ]; then
            break
        fi
        count=`expr $count + 1`
        sleep 1
        ocf_log debug "OpenStack Nova Scheduler (nova-scheduler) still" \
            "hasn't stopped yet. Waiting ..."
    done

    nova_scheduler_status
    rc=$?
    if [ $rc -ne $OCF_NOT_RUNNING ]; then
        # SIGTERM didn't help either, try SIGKILL
        ocf_log info "OpenStack Nova Scheduler (nova-scheduler) failed to" \
            "stop after ${shutdown_timeout}s using SIGTERM. Trying SIGKILL ..."
        ocf_run kill -s KILL $pid
    fi

    ocf_log info "OpenStack Nova Scheduler (nova-scheduler) 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_scheduler_validate || exit $?

# What kind of method was invoked?
case "$1" in
    start)
        nova_scheduler_start
        ;;
    stop)
        nova_scheduler_stop
        ;;
    status)
        nova_scheduler_status
        ;;
    monitor)
        nova_scheduler_monitor
        ;;
    validate-all)
        ;;
    *)
        usage
        exit $OCF_ERR_UNIMPLEMENTED
        ;;
esac