diff --git a/ocf/ceilometer-agent-central b/ocf/ceilometer-agent-central new file mode 100644 index 0000000..9c460a3 --- /dev/null +++ b/ocf/ceilometer-agent-central @@ -0,0 +1,345 @@ +#!/bin/sh +# +# +# OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) +# +# Description: Manages an OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) process as an HA resource +# +# Authors: Emilien Macchi +# Mainly inspired by the Nova Scheduler resource agent written by Sebastien Han +# +# 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_monitor_binary +# OCF_RESKEY_amqp_server_port +# OCF_RESKEY_additional_parameters +####################################################################### +# Initialization: + +: ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat} +. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs + +####################################################################### + +# Fill in some defaults if no values are specified + +OCF_RESKEY_binary_default="ceilometer-agent-central" +OCF_RESKEY_config_default="/etc/ceilometer/ceilometer.conf" +OCF_RESKEY_user_default="ceilometer" +OCF_RESKEY_pid_default="$HA_RSCTMP/$OCF_RESOURCE_INSTANCE.pid" +OCF_RESKEY_amqp_server_port_default="5672" + +: ${OCF_RESKEY_binary=${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}} + +####################################################################### + +usage() { + cat < + + +1.0 + + +Resource agent for the OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) +May manage a ceilometer-agent-central instance or a clone set that +creates a distributed ceilometer-agent-central cluster. + +Manages the OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) + + + + +Location of the OpenStack Ceilometer Central Agent server binary (ceilometer-agent-central) + +OpenStack Ceilometer Central Agent server binary (ceilometer-agent-central) + + + + + +Location of the OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) configuration file + +OpenStack Ceilometer Central Agent (ceilometer-agent-central registry) config file + + + + + +User running OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) + +OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) user + + + + + +The pid file to use for this OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) instance + +OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) pid file + + + + + +The listening port number of the AMQP server. Use for monitoring purposes + +AMQP listening port + + + + + + +Additional parameters to pass on to the OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) + +Additional parameters for ceilometer-agent-central + + + + + + + + + + + + + + +END +} + +####################################################################### +# Functions invoked by resource manager actions + +ceilometer_agent_central_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 +} + +ceilometer_agent_central_validate() { + local rc + + check_binary $OCF_RESKEY_binary + check_binary netstat + ceilometer_agent_central_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 +} + +ceilometer_agent_central_status() { + local pid + local rc + + if [ ! -f $OCF_RESKEY_pid ]; then + ocf_log info "OpenStack Ceilometer Central Agent (ceilometer-agent-central) 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 Ceilometer Central Agent (ceilometer-agent-central) is not running" + return $OCF_NOT_RUNNING + fi +} + +ceilometer_agent_central_monitor() { + local rc + local pid + local scheduler_amqp_check + + ceilometer_agent_central_status + rc=$? + + # 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 Cinder process with the same connection behavior (for example cinder-api) + pid=`cat $OCF_RESKEY_pid` + scheduler_amqp_check=`netstat -punt | grep -s "$OCF_RESKEY_amqp_server_port" | grep -s "$pid" | grep -qs "ESTABLISHED"` + rc=$? + if [ $rc -ne 0 ]; then + ocf_log err "Central Agent is not connected to the AMQP server : $rc" + return $OCF_NOT_RUNNING + fi + + ocf_log debug "OpenStack Ceilometer Central Agent (ceilometer-agent-central) monitor succeeded" + return $OCF_SUCCESS +} + +ceilometer_agent_central_start() { + local rc + + ceilometer_agent_central_status + rc=$? + if [ $rc -eq $OCF_SUCCESS ]; then + ocf_log info "OpenStack Ceilometer Central Agent (ceilometer-agent-central) already running" + return $OCF_SUCCESS + fi + + # run the actual ceilometer-agent-central 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 $!' > $OCF_RESKEY_pid + + # Spin waiting for the server to come up. + while true; do + ceilometer_agent_central_monitor + rc=$? + [ $rc -eq $OCF_SUCCESS ] && break + if [ $rc -ne $OCF_NOT_RUNNING ]; then + ocf_log err "OpenStack Ceilometer Central Agent (ceilometer-agent-central) start failed" + exit $OCF_ERR_GENERIC + fi + sleep 1 + done + + ocf_log info "OpenStack Ceilometer Central Agent (ceilometer-agent-central) started" + return $OCF_SUCCESS +} + +ceilometer_agent_central_stop() { + local rc + local pid + + ceilometer_agent_central_status + rc=$? + if [ $rc -eq $OCF_NOT_RUNNING ]; then + ocf_log info "OpenStack Ceilometer Central Agent (ceilometer-agent-central) 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 Ceilometer Central Agent (ceilometer-agent-central) 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 $shutdown_timeout ]; do + ceilometer_agent_central_status + rc=$? + if [ $rc -eq $OCF_NOT_RUNNING ]; then + break + fi + count=`expr $count + 1` + sleep 1 + ocf_log debug "OpenStack Ceilometer Central Agent (ceilometer-agent-central) still hasn't stopped yet. Waiting ..." + done + + ceilometer_agent_central_status + rc=$? + if [ $rc -ne $OCF_NOT_RUNNING ]; then + # SIGTERM didn't help either, try SIGKILL + ocf_log info "OpenStack Ceilometer Central Agent (ceilometer-agent-central) failed to stop after ${shutdown_timeout}s \ + using SIGTERM. Trying SIGKILL ..." + ocf_run kill -s KILL $pid + fi + + ocf_log info "OpenStack Ceilometer Central Agent (ceilometer-agent-central) 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 +ceilometer_agent_central_validate || exit $? + +# What kind of method was invoked? +case "$1" in + start) ceilometer_agent_central_start;; + stop) ceilometer_agent_central_stop;; + status) ceilometer_agent_central_status;; + monitor) ceilometer_agent_central_monitor;; + validate-all) ;; + *) usage + exit $OCF_ERR_UNIMPLEMENTED;; +esac