stackforge/puppet-openstack-cloud
Code Issues Proposed changes

ceilometer-central-agent: Remove dependency from pacemaker

Browse Source 
Remove all related configuration that sets up ceilometer-central-agent
and pacemaker. ceilometer-central-agent will now use tooz w/
coordination url and a Redis cluster to be active/active.

Change-Id: I3afbcca43b243844fef9602693d6e86ff9844a59
This commit is contained in:
Yanis Guenane
2015-03-02 16:54:22 +01:00
parent d32ee81ca7
commit f46d8d58ad
6 changed files with 4 additions and 447 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
files/heartbeat/README.md
View File

@@ -1,3 +0,0 @@
# Pacemaker High Availability resource agents for OpenStack
https://github.com/madkiss/openstack-resource-agents

345
files/heartbeat/ceilometer-agent-central
View File

@@ -1,345 +0,0 @@
#!/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 <<UEND
usage: $0 (start|stop|validate-all|meta-data|status|monitor)
$0 manages an OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) process as an HA resource
The 'start' operation starts the scheduler service.
The 'stop' operation stops the scheduler service.
The 'validate-all' operation reports whether the parameters are valid
The 'meta-data' operation reports this RA's meta-data information
The 'status' operation reports whether the scheduler service is running
The 'monitor' operation reports whether the scheduler service seems to be working
UEND
}
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="ceilometer-agent-central">
<version>1.0</version>
<longdesc lang="en">
Resource agent for the OpenStack Ceilometer Central Agent Service (ceilometer-agent-central)
May manage a ceilometer-agent-central instance or a clone set that
creates a distributed ceilometer-agent-central cluster.
</longdesc>
<shortdesc lang="en">Manages the OpenStack Ceilometer Central Agent Service (ceilometer-agent-central)</shortdesc>
<parameters>
<parameter name="binary" unique="0" required="0">
<longdesc lang="en">
Location of the OpenStack Ceilometer Central Agent server binary (ceilometer-agent-central)
</longdesc>
<shortdesc lang="en">OpenStack Ceilometer Central Agent server binary (ceilometer-agent-central)</shortdesc>
<content type="string" default="${OCF_RESKEY_binary_default}" />
</parameter>
<parameter name="config" unique="0" required="0">
<longdesc lang="en">
Location of the OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) configuration file
</longdesc>
<shortdesc lang="en">OpenStack Ceilometer Central Agent (ceilometer-agent-central registry) config file</shortdesc>
<content type="string" default="${OCF_RESKEY_config_default}" />
</parameter>
<parameter name="user" unique="0" required="0">
<longdesc lang="en">
User running OpenStack Ceilometer Central Agent Service (ceilometer-agent-central)
</longdesc>
<shortdesc lang="en">OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) user</shortdesc>
<content type="string" default="${OCF_RESKEY_user_default}" />
</parameter>
<parameter name="pid" unique="0" required="0">
<longdesc lang="en">
The pid file to use for this OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) instance
</longdesc>
<shortdesc lang="en">OpenStack Ceilometer Central Agent Service (ceilometer-agent-central) pid file</shortdesc>
<content type="string" default="${OCF_RESKEY_pid_default}" />
</parameter>
<parameter name="amqp_server_port" unique="0" required="0">
<longdesc lang="en">
The listening port number of the AMQP server. Use for monitoring purposes
</longdesc>
<shortdesc lang="en">AMQP listening port</shortdesc>
<content type="integer" default="${OCF_RESKEY_amqp_server_port_default}" />
</parameter>
<parameter name="additional_parameters" unique="0" required="0">
<longdesc lang="en">
Additional parameters to pass on to the OpenStack Ceilometer Central Agent Service (ceilometer-agent-central)
</longdesc>
<shortdesc lang="en">Additional parameters for ceilometer-agent-central</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
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

35
manifests/spof.pp
View File

@@ -66,7 +66,6 @@ class cloud::spof(
class {'pacemaker::stonith':
disable => true
}
pacemaker::resource::systemd { 'openstack-ceilometer-central': }
} else {
class { 'corosync':
@@ -87,43 +86,9 @@ class cloud::spof(
'pe-warn-series-max': value => 1000;
'pe-input-series-max': value => 1000;
'cluster-recheck-interval': value => '5min';
} ->
file { '/usr/lib/ocf/resource.d/heartbeat/ceilometer-agent-central':
source => 'puppet:///modules/cloud/heartbeat/ceilometer-agent-central',
mode => '0755',
owner => 'root',
group => 'root',
} ->
cs_primitive { 'ceilometer-agent-central':
primitive_class => 'ocf',
primitive_type => 'ceilometer-agent-central',
provided_by => 'heartbeat',
operations => {
'monitor' => {
interval => '10s',
timeout => '30s'
},
'start' => {
interval => '0',
timeout => '30s',
on-fail => 'restart'
}
}
} ->
exec { 'cleanup_ceilometer_agent_central':
command => 'crm resource cleanup ceilometer-agent-central',
unless => 'crm resource show ceilometer-agent-central | grep Started',
user => 'root',
path => ['/usr/sbin', '/bin'],
}
}
# Run OpenStack SPOF service and disable them since they will be managed by Corosync.
class { 'cloud::telemetry::centralagent':
enabled => false,
}
if $::cloud::manage_firewall {
cloud::firewall::rule{ '100 allow vrrp access':
port => undef,

24
manifests/telemetry/centralagent.pp
View File

@@ -16,29 +16,11 @@
#
# == Class: cloud::telemetry::centralagent
#
# Telemetry Central Agent node (should be run once)
# Could be managed by spof node as Active / Passive.
# Telemetry Central Agent node
#
# === Parameters:
#
# [*enabled*]
# (optional) State of the telemetry central agent service.
# Defaults to true
#
# [*coordination_url*]
# (optional) The url to use for distributed group membership coordination.
# Defaults to undef
#
class cloud::telemetry::centralagent(
$enabled = true,
$coordination_url = undef,
){
class cloud::telemetry::centralagent{
include 'cloud::telemetry'
class { 'ceilometer::agent::central':
enabled => $enabled,
coordination_url => $coordination_url,
}
include 'ceilometer::agent::central'
}

15
spec/classes/cloud_spof_spec.rb
View File

@@ -41,19 +41,6 @@ describe 'cloud::spof' do
:bind_address => '10.0.0.1',
:multicast_address => '239.1.1.2',
)
is_expected.to contain_file('/usr/lib/ocf/resource.d/heartbeat/ceilometer-agent-central').with(
:source => 'puppet:///modules/cloud/heartbeat/ceilometer-agent-central',
:mode => '0755',
:owner => 'root',
:group => 'root'
)
is_expected.to contain_class('cloud::telemetry::centralagent').with(:enabled => false)
is_expected.to contain_exec('cleanup_ceilometer_agent_central').with(
:command => 'crm resource cleanup ceilometer-agent-central',
:path => ['/usr/sbin', '/bin'],
:user => 'root',
:unless => 'crm resource show ceilometer-agent-central | grep Started'
)
end
end
@@ -73,8 +60,6 @@ describe 'cloud::spof' do
:settle_try_sleep => 5,
:manage_fw => false,
:cluster_members => 'srv1 srv2 srv3')
is_expected.to contain_pcmk_resource('openstack-ceilometer-central')
is_expected.to contain_class('cloud::telemetry::centralagent').with(:enabled => false)
end
end

29
spec/classes/cloud_telemetry_centralagent_spec.rb
View File

@@ -38,35 +38,8 @@ describe 'cloud::telemetry::centralagent' do
debug => true }"
end
let :params do
{ :enabled => 'true',
:coordination_url => 'http://coordination' }
end
it 'configure ceilometer common' do
is_expected.to contain_class('ceilometer').with(
:verbose => true,
:debug => true,
:rabbit_userid => 'ceilometer',
:rabbit_hosts => ['10.0.0.1'],
:rabbit_password => 'secrete',
:metering_secret => 'secrete',
:use_syslog => true,
:log_facility => 'LOG_LOCAL0',
:log_dir => false
)
is_expected.to contain_class('ceilometer::agent::auth').with(
:auth_password => 'secrete',
:auth_url => 'http://10.0.0.1:5000/v2.0',
:auth_region => 'MyRegion',
)
end
it 'configure ceilometer central agent' do
is_expected.to contain_class('ceilometer::agent::central').with({
'enabled' => 'true',
'coordination_url'=> 'http://coordination',
})
is_expected.to contain_class('ceilometer::agent::central')
end
end