ceilometer-agent-central has been replaced by ceilometer-polling

The ceilometer-agent-central and ceilometer-agent-compute
services are now completely removed from the liberty release
of ceilometer. The new implementation is to use
ceilometer-polling and call the respective pollsters by
namespace.

Change-Id: I3af6fe9cc9c5b0fcf0406f89f402cbd1a31f66c4
Closes-Bug: 1606787
This commit is contained in:
Zhang Hua 2016-08-26 10:20:58 +08:00 committed by Edward Hope-Morley
parent ec93c4eb56
commit 3816946e2f
4 changed files with 421 additions and 11 deletions

View File

@ -43,6 +43,7 @@ from charmhelpers.core.host import (
)
from charmhelpers.contrib.openstack.utils import (
configure_installation_source,
os_release,
openstack_upgrade_available,
pausable_restart_on_change as restart_on_change,
is_unit_paused_set,
@ -90,6 +91,10 @@ from charmhelpers.contrib.hardening.harden import harden
hooks = Hooks()
CONFIGS = register_configs()
LEGACY_RES_KEY = 'res_ceilometer_agent_central'
LEGACY_OCF_FILE = "/usr/lib/ocf/resource.d/openstack/ceilometer-agent-central"
NEW_RES_KEY = 'res_ceilometer_polling'
@hooks.hook('install.real')
@harden()
@ -189,6 +194,15 @@ def config_changed():
for rid in relation_ids('identity-service'):
keystone_joined(relid=rid)
# Install OCF resource file locally
install_ceilometer_ocf()
# Define the new ocf resource and use the key delete_resources to delete
# legacy resource for >= Liberty since the ceilometer-agent-central moved
# to ceilometer-polling in liberty (see LP: #1606787).
for rid in relation_ids('ha'):
ha_joined(rid)
@hooks.hook('upgrade-charm')
@harden()
@ -199,13 +213,21 @@ def upgrade_charm():
def install_ceilometer_ocf():
dest_file = "/usr/lib/ocf/resource.d/openstack/ceilometer-agent-central"
dest_file = LEGACY_OCF_FILE
src_file = 'ocf/openstack/ceilometer-agent-central'
if os_release('ceilometer-common') >= 'liberty':
dest_file = "/usr/lib/ocf/resource.d/openstack/ceilometer-polling"
src_file = 'ocf/openstack/ceilometer-polling'
# delete ocf file due to package no longer exists after upgrade
if os.path.isfile(LEGACY_OCF_FILE):
os.remove(LEGACY_OCF_FILE)
if not os.path.isdir(os.path.dirname(dest_file)):
os.makedirs(os.path.dirname(dest_file))
if not os.path.exists(dest_file):
shutil.copy(src_file, dest_file)
os.chmod(dest_file, 0o755)
@hooks.hook('cluster-relation-joined')
@ -236,16 +258,22 @@ def cluster_changed():
@hooks.hook('ha-relation-joined')
def ha_joined(relation_id=None):
cluster_config = get_hacluster_config()
delete_resources = []
RES_KEY = LEGACY_RES_KEY
if os_release('ceilometer-common') >= 'liberty':
RES_KEY = NEW_RES_KEY
delete_resources.append(LEGACY_RES_KEY)
ocf_name = RES_KEY.replace('res_', '').replace('_', '-')
resources = {
'res_ceilometer_haproxy': 'lsb:haproxy',
'res_ceilometer_agent_central': ('ocf:openstack:'
'ceilometer-agent-central')
RES_KEY: ('ocf:openstack:{}'.format(ocf_name))
}
resource_params = {
'res_ceilometer_haproxy': 'op monitor interval="5s"',
'res_ceilometer_agent_central': 'op monitor interval="30s"'
RES_KEY: 'op monitor interval="30s"'
}
amqp_ssl_port = None
@ -256,7 +284,7 @@ def ha_joined(relation_id=None):
if amqp_ssl_port:
params = ('params amqp_server_port="%s" op monitor interval="30s"' %
(amqp_ssl_port))
resource_params['res_ceilometer_agent_central'] = params
resource_params[RES_KEY] = params
if config('dns-ha'):
update_dns_ha_resource_params(relation_id=relation_id,
@ -283,7 +311,9 @@ def ha_joined(relation_id=None):
vip_group.append(vip_key)
if len(vip_group) >= 1:
relation_set(groups={'grp_ceilometer_vips': ' '.join(vip_group)})
relation_set(relation_id=relation_id,
groups={'grp_ceilometer_vips':
' '.join(vip_group)})
init_services = {
'res_ceilometer_haproxy': 'haproxy'
@ -297,6 +327,7 @@ def ha_joined(relation_id=None):
corosync_mcastport=cluster_config['ha-mcastport'],
resources=resources,
resource_params=resource_params,
delete_resources=delete_resources,
clones=clones)

View File

@ -240,6 +240,8 @@ ceilometer_agent_central_monitor() {
ceilometer_agent_central_start() {
local rc
service ceilometer-agent-central stop >> /dev/null 2>&1
ceilometer_agent_central_status
rc=$?
if [ $rc -eq $OCF_SUCCESS ]; then

View File

@ -0,0 +1,348 @@
#!/bin/sh
#
#
# OpenStack Ceilometer Polling Agent Service (ceilometer-polling)
#
# Description: Manages an OpenStack Ceilometer Polling Agent Service (ceilometer-polling) 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}
. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs
#######################################################################
# Fill in some defaults if no values are specified
OCF_RESKEY_binary_default="ceilometer-polling"
OCF_RESKEY_additional_parameters=" --polling-namespaces central"
OCF_RESKEY_config_default="/etc/ceilometer/ceilometer.conf"
OCF_RESKEY_user_default="ceilometer"
OCF_RESKEY_pid_default="$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)
$0 manages an OpenStack Ceilometer Polling Agent Service (ceilometer-polling) process as an HA resource
The 'start' operation starts the scheduler service.
The 'stop' operation stops the scheduler service.
The 'validate-all' operation reports whether the parameters are valid
The 'meta-data' operation reports this RA's meta-data information
The 'status' operation reports whether the scheduler service is running
The 'monitor' operation reports whether the scheduler service seems to be working
UEND
}
meta_data() {
cat <<END
<?xml version="1.0"?>
<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
<resource-agent name="ceilometer-polling">
<version>1.0</version>
<longdesc lang="en">
Resource agent for the OpenStack Ceilometer Polling Agent Service (ceilometer-polling)
May manage a ceilometer-polling instance or a clone set that
creates a distributed ceilometer-polling cluster.
</longdesc>
<shortdesc lang="en">Manages the OpenStack Ceilometer Polling Agent Service (ceilometer-polling)</shortdesc>
<parameters>
<parameter name="binary" unique="0" required="0">
<longdesc lang="en">
Location of the OpenStack Ceilometer Polling Agent server binary (ceilometer-polling)
</longdesc>
<shortdesc lang="en">OpenStack Ceilometer Polling Agent server binary (ceilometer-polling)</shortdesc>
<content type="string" default="${OCF_RESKEY_binary_default}" />
</parameter>
<parameter name="config" unique="0" required="0">
<longdesc lang="en">
Location of the OpenStack Ceilometer Polling Agent Service (ceilometer-polling) configuration file
</longdesc>
<shortdesc lang="en">OpenStack Ceilometer Polling Agent (ceilometer-polling registry) config file</shortdesc>
<content type="string" default="${OCF_RESKEY_config_default}" />
</parameter>
<parameter name="user" unique="0" required="0">
<longdesc lang="en">
User running OpenStack Ceilometer Polling Agent Service (ceilometer-polling)
</longdesc>
<shortdesc lang="en">OpenStack Ceilometer Polling Agent Service (ceilometer-polling) user</shortdesc>
<content type="string" default="${OCF_RESKEY_user_default}" />
</parameter>
<parameter name="pid" unique="0" required="0">
<longdesc lang="en">
The pid file to use for this OpenStack Ceilometer Polling Agent Service (ceilometer-polling) instance
</longdesc>
<shortdesc lang="en">OpenStack Ceilometer Polling Agent Service (ceilometer-polling) 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 Polling Agent Service (ceilometer-polling)
</longdesc>
<shortdesc lang="en">Additional parameters for ceilometer-polling</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_polling_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_polling_validate() {
local rc
check_binary $OCF_RESKEY_binary
check_binary netstat
ceilometer_polling_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_polling_status() {
local pid
local rc
if [ ! -f $OCF_RESKEY_pid ]; then
ocf_log info "OpenStack Ceilometer Polling Agent (ceilometer-polling) 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 Polling Agent (ceilometer-polling) is not running"
return $OCF_NOT_RUNNING
fi
}
ceilometer_polling_monitor() {
local rc
local pid
local scheduler_amqp_check
ceilometer_polling_status
rc=$?
# 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 Polling Agent (ceilometer-polling) monitor succeeded"
return $OCF_SUCCESS
}
ceilometer_polling_start() {
local rc
service ceilometer-agent-central stop >> /dev/null 2>&1
ceilometer_polling_status
rc=$?
if [ $rc -eq $OCF_SUCCESS ]; then
ocf_log info "OpenStack Ceilometer Polling Agent (ceilometer-polling) already running"
return $OCF_SUCCESS
fi
# run the actual ceilometer-polling 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 $!' > $OCF_RESKEY_pid
# Spin waiting for the server to come up.
while true; do
ceilometer_polling_monitor
rc=$?
[ $rc -eq $OCF_SUCCESS ] && break
if [ $rc -ne $OCF_NOT_RUNNING ]; then
ocf_log err "OpenStack Ceilometer Polling Agent (ceilometer-polling) start failed"
exit $OCF_ERR_GENERIC
fi
sleep 1
done
ocf_log info "OpenStack Ceilometer Polling Agent (ceilometer-polling) started"
return $OCF_SUCCESS
}
ceilometer_polling_stop() {
local rc
local pid
ceilometer_polling_status
rc=$?
if [ $rc -eq $OCF_NOT_RUNNING ]; then
ocf_log info "OpenStack Ceilometer Polling Agent (ceilometer-polling) 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 Polling Agent (ceilometer-polling) 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 $shutdown_timeout ]; do
ceilometer_polling_status
rc=$?
if [ $rc -eq $OCF_NOT_RUNNING ]; then
break
fi
count=`expr $count + 1`
sleep 1
ocf_log debug "OpenStack Ceilometer Polling Agent (ceilometer-polling) still hasn't stopped yet. Waiting ..."
done
ceilometer_polling_status
rc=$?
if [ $rc -ne $OCF_NOT_RUNNING ]; then
# SIGTERM didn't help either, try SIGKILL
ocf_log info "OpenStack Ceilometer Polling Agent (ceilometer-polling) failed to stop after ${shutdown_timeout}s \
using SIGTERM. Trying SIGKILL ..."
ocf_run kill -s KILL $pid
fi
ocf_log info "OpenStack Ceilometer Polling Agent (ceilometer-polling) 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_polling_validate || exit $?
# What kind of method was invoked?
case "$1" in
start) ceilometer_polling_start;;
stop) ceilometer_polling_stop;;
status) ceilometer_polling_status;;
monitor) ceilometer_polling_monitor;;
validate-all) ;;
*) usage
exit $OCF_ERR_UNIMPLEMENTED;;
esac

View File

@ -66,6 +66,7 @@ TO_PATCH = [
'reload_systemd',
'mkdir',
'init_is_systemd',
'os_release',
]
@ -148,7 +149,9 @@ class CeilometerHooksTest(CharmTestCase):
@patch('charmhelpers.core.hookenv.config')
@patch.object(hooks, 'ceilometer_joined')
def test_config_changed_no_upgrade(self, joined, mock_config):
@patch.object(hooks, 'install_ceilometer_ocf')
def test_config_changed_no_upgrade(self, ocf,
joined, mock_config):
self.openstack_upgrade_available.return_value = False
hooks.hooks.execute(['hooks/config-changed'])
self.openstack_upgrade_available.\
@ -157,10 +160,13 @@ class CeilometerHooksTest(CharmTestCase):
self.assertTrue(self.CONFIGS.write_all.called)
self.assertTrue(joined.called)
self.assertTrue(self.reload_systemd.called)
self.assertTrue(ocf.called)
@patch('charmhelpers.core.hookenv.config')
@patch.object(hooks, 'ceilometer_joined')
def test_config_changed_upgrade(self, joined, mock_config):
@patch.object(hooks, 'install_ceilometer_ocf')
def test_config_changed_upgrade(self, ocf,
joined, mock_config):
self.openstack_upgrade_available.return_value = True
hooks.hooks.execute(['hooks/config-changed'])
self.openstack_upgrade_available.\
@ -169,14 +175,17 @@ class CeilometerHooksTest(CharmTestCase):
self.assertTrue(self.CONFIGS.write_all.called)
self.assertTrue(joined.called)
self.assertTrue(self.reload_systemd.called)
self.assertTrue(ocf.called)
def test_config_changed_with_openstack_upgrade_action(self):
@patch.object(hooks, 'install_ceilometer_ocf')
def test_config_changed_with_openstack_upgrade_action(self, ocf):
self.openstack_upgrade_available.return_value = True
self.test_config.set('action-managed-upgrade', True)
hooks.hooks.execute(['hooks/config-changed'])
self.assertFalse(self.do_openstack_upgrade.called)
self.assertTrue(ocf.called)
@patch.object(hooks, 'canonical_url')
@patch('charmhelpers.core.hookenv.config')
@ -296,6 +305,7 @@ class CeilometerHooksTest(CharmTestCase):
@patch.object(hooks, 'get_netmask_for_address')
def test_ha_joined(self, mock_netmask, mock_iface, mock_cluster_config,
mock_config):
self.os_release.return_value = 'kilo'
mock_cluster_config.return_value = {'vip': '10.0.5.100',
'ha-bindiface': 'bnd0',
'ha-mcastport': 5802}
@ -318,13 +328,15 @@ class CeilometerHooksTest(CharmTestCase):
'nic="eth0"')
}
exp_clones = {'cl_ceilometer_haproxy': 'res_ceilometer_haproxy'}
call1 = call(groups={'grp_ceilometer_vips': 'res_ceilometer_eth0_vip'})
call1 = call(relation_id=None,
groups={'grp_ceilometer_vips': 'res_ceilometer_eth0_vip'})
call2 = call(relation_id=None,
init_services={'res_ceilometer_haproxy': 'haproxy'},
corosync_bindiface='bnd0',
corosync_mcastport=5802,
resources=exp_resources,
resource_params=exp_resource_params,
delete_resources=[],
clones=exp_clones)
self.relation_set.assert_has_calls([call1, call2], any_order=False)
@ -338,6 +350,7 @@ class CeilometerHooksTest(CharmTestCase):
def test_ha_joined_ssl(self, mock_rel_get, mock_rel_units, mock_rel_ids,
mock_iface, mock_cluster_config, mock_netmask,
mock_config):
self.os_release.return_value = 'kilo'
mock_rel_ids.return_value = 'amqp:0'
mock_rel_units.return_value = 'rabbitmq-server/0'
mock_rel_get.return_value = '5671'
@ -366,13 +379,15 @@ class CeilometerHooksTest(CharmTestCase):
'nic="eth0"')
}
exp_clones = {'cl_ceilometer_haproxy': 'res_ceilometer_haproxy'}
call1 = call(groups={'grp_ceilometer_vips': 'res_ceilometer_eth0_vip'})
call1 = call(relation_id=None,
groups={'grp_ceilometer_vips': 'res_ceilometer_eth0_vip'})
call2 = call(relation_id=None,
init_services={'res_ceilometer_haproxy': 'haproxy'},
corosync_bindiface='bnd0',
corosync_mcastport=5802,
resources=exp_resources,
resource_params=exp_resource_params,
delete_resources=[],
clones=exp_clones)
self.relation_set.assert_has_calls([call1, call2], any_order=False)
@ -386,6 +401,7 @@ class CeilometerHooksTest(CharmTestCase):
'ip_address="10.0.0.1"'})
self.test_config.set('dns-ha', True)
self.os_release.return_value = 'kilo'
mock_cluster_config.return_value = {
'vip': None,
'ha-bindiface': 'em0',
@ -409,6 +425,7 @@ class CeilometerHooksTest(CharmTestCase):
'ip_address="10.0.0.1"',
'res_ceilometer_haproxy': 'op monitor interval="5s"',
'res_ceilometer_agent_central': 'op monitor interval="30s"'},
'delete_resources': [],
'clones': {'cl_ceilometer_haproxy': 'res_ceilometer_haproxy'}
}
self.update_dns_ha_resource_params.side_effect = _fake_update
@ -417,6 +434,18 @@ class CeilometerHooksTest(CharmTestCase):
self.assertTrue(self.update_dns_ha_resource_params.called)
self.relation_set.assert_called_with(**args)
self.os_release.return_value = 'liberty'
args.get('resources').pop('res_ceilometer_agent_central')
args.get('resource_params').pop('res_ceilometer_agent_central')
args.get('resources')['res_ceilometer_polling'] = \
'ocf:openstack:ceilometer-polling'
args.get('resource_params')['res_ceilometer_polling'] = \
'op monitor interval="30s"'
args['delete_resources'] = ['res_ceilometer_agent_central']
self.update_dns_ha_resource_params.side_effect = _fake_update
hooks.ha_joined()
self.relation_set.assert_called_with(**args)
@patch('charmhelpers.core.hookenv.config')
@patch.object(hooks, 'keystone_joined')
def test_ha_changed_not_clustered(self, mock_keystone_joined, mock_config):