x/fuel-plugin-lma-collector
Code Issues Proposed changes

Update service status filter for InfluxDB 0.9

Browse Source 
This change modifies the service accumulator and global service status
filters to support the new data model implemented for InfluxDB 0.9.

Change-Id: Id437d1111660d2d55bb7ba8dc46ed8c8eb8e41b7
Implements: blueprint upgrade-influxdb-grafana
This commit is contained in:
Simon Pasquier
2015-08-05 10:50:33 +02:00
parent 8f268f04ff
commit 098c3189f5
9 changed files with 585 additions and 93 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
deployment_scripts/puppet/manifests/controller.pp
View File

@@ -162,7 +162,12 @@ if $lma_collector['influxdb_mode'] != 'disabled' {
}
# Service status metrics and annotations
class { 'lma_collector::metrics::service_status': }
if $lma_collector['influxdb_legacy'] {
class { 'lma_collector::metrics::service_status_legacy': }
} else {
class { 'lma_collector::metrics::service_status': }
}
}
$alerting_mode = $lma_collector['alerting_mode']

9
deployment_scripts/puppet/modules/lma_collector/files/plugins/decoders/collectd.lua
View File

@@ -157,6 +157,11 @@ function process_message ()
elseif metric_source == 'check_openstack_api' then
-- For OpenStack API metrics, plugin_instance = <service name>
msg['Fields']['name'] = 'openstack' .. sep .. sample['plugin_instance'] .. sep .. 'check_api'
-- 'service' isn't used as a tag because there is no point to
-- aggregate data across all services. It is stored in the
-- Fields table though because it simplifies the life of
-- downstream filters consuming this data.
msg['Fields']['service'] = sample['plugin_instance']
if sample['type_instance'] ~= nil and sample['type_instance'] ~= '' then
msg['Fields']['os_region'] = sample['type_instance']
end
@@ -270,7 +275,9 @@ function process_message ()
msg['Fields']['service'] = service
msg['Fields']['state'] = state
elseif metric_source == 'pacemaker_resource' then
msg['Fields']['name'] = 'pacemaker_resource' .. sep .. sample['type_instance'] .. sep .. 'active'
msg['Fields']['name'] = 'pacemaker_local_resource_active'
msg['Fields']['tag_fields'] = { 'resource' }
msg['Fields']['resource'] = sample['type_instance']
elseif metric_source == 'users' then
-- 'users' is a reserved name for InfluxDB v0.9
msg['Fields']['name'] = 'logged_users'

64
deployment_scripts/puppet/modules/lma_collector/files/plugins/filters/service_accumulator_states.lua
View File

@@ -77,24 +77,29 @@ require 'math'
local floor = math.floor
local utils = require 'lma_utils'
_PRESERVATION_VERSION = 1
_PRESERVATION_VERSION = 2
-- variables with global scope are preserved between restarts
services = {}
vip_active_at = 0
local payload_name = read_config('inject_payload_name') or 'service_status'
local state_label_map = {
up = utils.state_map.UP,
down = utils.state_map.DOWN,
disabled = utils.state_map.DISABLED,
}
function process_message ()
local ts = floor(read_message("Timestamp")/1e6) -- ms
local ts = read_message("Timestamp")
local metric_name = read_message("Fields[name]")
local value = read_message("Fields[value]")
local state = state_label_map[read_message('Fields[state]')]
local name
local top_entry
local item_name
local group_name
local state
if string.find(metric_name, '^pacemaker.resource.vip__public') then
if metric_name == 'pacemaker_local_resource_active' and read_message("Fields[resource]") == 'vip__public' then
if value == 1 then
vip_active_at = ts
else
@@ -103,42 +108,34 @@ function process_message ()
return 0
end
if string.find(metric_name, '%.up$') then
state = utils.state_map.UP
elseif string.find(metric_name, '%.down$') then
state = utils.state_map.DOWN
elseif string.find(metric_name, '%.disabled$') then
state = utils.state_map.DISABLED
end
if string.find(metric_name, '^openstack') then
name, group_name, item_name = string.match(metric_name, '^openstack%.([^._]+)%.([^._]+)%.([^._]+)')
if string.find(metric_name, '^openstack_([^._]+)_services$') or string.find(metric_name, '^openstack_([^._]+)_agents$') then
name, group_name = string.match(metric_name, '([^_]+)_([^_]+)$')
top_entry = 'workers'
if not item_name then
-- A service can have several API checks, by convention the service name
-- is written down "<name>-<item>" or just "<name>".
item_name = string.match(metric_name, '^openstack%.([^.]+)%.check_api$')
name, _ = string.match(item_name, '^([^-]+)\-(.*)')
if not name then
name = item_name
end
item_name = read_message('Fields[service]')
top_entry = 'check_api'
group_name = 'endpoint'
-- retrieve the current state
state = utils.check_api_status_to_state_map[value]
-- and always override value to 1
value = 1
end
elseif string.find(metric_name, '_check_api$') then
-- A service can have several API checks, by convention the service name
-- is written down "<name>-<item>" or just "<name>".
name = string.match(read_message('Fields[service]'), '^([^-]+)')
top_entry = 'check_api'
group_name = 'endpoint'
item_name = read_message('Fields[service]')
elseif string.find(metric_name, '^haproxy%.backend') then
-- convert 0/1 value to up/down state
state = utils.check_api_status_to_state_map[value]
-- and always override value to 1
value = 1
elseif metric_name == 'haproxy_backend_servers' then
name = string.match(read_message('Fields[backend]'), '^([^-]+)')
top_entry = 'haproxy'
group_name = 'pool'
item_name = string.match(metric_name, '^haproxy%.backend%.([^.]+)%.servers')
name = string.match(item_name, '^([^-]+)')
item_name = read_message('Fields[backend]')
end
if not name or not item_name then
return -1
return -1
end
-- table initialization for the first time we see a service
@@ -165,6 +162,7 @@ function process_message ()
service[invert_state] = {last_seen=ts, value=0, group_name=group_name}
end
end
return 0
end

176
deployment_scripts/puppet/modules/lma_collector/files/plugins/filters/service_accumulator_states_legacy.lua Normal file
View File

@@ -0,0 +1,176 @@
-- Copyright 2015 Mirantis, Inc.
--
-- Licensed under the Apache License, Version 2.0 (the "License");
-- you may not use this file except in compliance with the License.
-- You may obtain a copy of the License at
--
-- http://www.apache.org/licenses/LICENSE-2.0
--
-- Unless required by applicable law or agreed to in writing, software
-- distributed under the License is distributed on an "AS IS" BASIS,
-- WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-- See the License for the specific language governing permissions and
-- limitations under the License.
-- The filter accumulates data into a table and emits regularly a message per
-- service with a payload like this:
-- {
-- "vip_active_at": 1435829917607,
-- "name": "nova",
-- "states": {
-- "check_api":{
-- "nova":{
-- "down":{
-- "value":0,
-- "group_name":"endpoint",
-- "last_seen":1433252000524
-- },
-- "up":{
-- "value":1,
-- "group_name":"endpoint",
-- "last_seen":1433252000524
-- }
-- },
-- ...
-- },
-- "workers":{
-- "scheduler":{
-- "down":{
-- "value":0,
-- "group_name":"services",
-- "last_seen":1433251999229
-- },
-- "disabled":{
-- "value":1,
-- "group_name":"services",
-- "last_seen":1433251999226
-- },
-- "up":{
-- "value":2,
-- "group_name":"services",
-- "last_seen":1433251999227
-- }
-- },
-- ...
-- },
-- "haproxy":{
-- "nova-api":{
-- "down":{
-- "value":0,
-- "group_name":"pool",
-- "last_seen":1433252000957
-- },
-- "up":{
-- "value":3,
-- "group_name":"pool",
-- "last_seen":1433252000954
-- }
-- }
-- }
-- ...
-- }
-- }
require 'cjson'
require 'string'
require 'math'
local floor = math.floor
local utils = require 'lma_utils'
_PRESERVATION_VERSION = 1
-- variables with global scope are preserved between restarts
services = {}
vip_active_at = 0
local payload_name = read_config('inject_payload_name') or 'service_status'
function process_message ()
local ts = floor(read_message("Timestamp")/1e6) -- ms
local metric_name = read_message("Fields[name]")
local value = read_message("Fields[value]")
local name
local top_entry
local item_name
local group_name
local state
if string.find(metric_name, '^pacemaker.resource.vip__public') then
if value == 1 then
vip_active_at = ts
else
vip_active_at = 0
end
return 0
end
if string.find(metric_name, '%.up$') then
state = utils.state_map.UP
elseif string.find(metric_name, '%.down$') then
state = utils.state_map.DOWN
elseif string.find(metric_name, '%.disabled$') then
state = utils.state_map.DISABLED
end
if string.find(metric_name, '^openstack') then
name, group_name, item_name = string.match(metric_name, '^openstack%.([^._]+)%.([^._]+)%.([^._]+)')
top_entry = 'workers'
if not item_name then
-- A service can have several API checks, by convention the service name
-- is written down "<name>-<item>" or just "<name>".
item_name = string.match(metric_name, '^openstack%.([^.]+)%.check_api$')
name, _ = string.match(item_name, '^([^-]+)\-(.*)')
if not name then
name = item_name
end
top_entry = 'check_api'
group_name = 'endpoint'
-- retrieve the current state
state = utils.check_api_status_to_state_map[value]
-- and always override value to 1
value = 1
end
elseif string.find(metric_name, '^haproxy%.backend') then
top_entry = 'haproxy'
group_name = 'pool'
item_name = string.match(metric_name, '^haproxy%.backend%.([^.]+)%.servers')
name = string.match(item_name, '^([^-]+)')
end
if not name or not item_name then
return -1
end
-- table initialization for the first time we see a service
if not services[name] then services[name] = {} end
if not services[name][top_entry] then services[name][top_entry] = {} end
if not services[name][top_entry][item_name] then services[name][top_entry][item_name] = {} end
local service = services[name][top_entry][item_name]
service[state] = {last_seen=ts, value=value, group_name=group_name}
-- In the logic to treat check_api results like others, group by up/down
-- and reset the counterpart w/ value=0
if top_entry == 'check_api' then
local invert_state
if state == utils.state_map.UP then
invert_state = utils.state_map.DOWN
elseif state == utils.state_map.DOWN then
invert_state = utils.state_map.UP
end
if invert_state then
if not service[invert_state] then
service[invert_state] = {}
end
service[invert_state] = {last_seen=ts, value=0, group_name=group_name}
end
end
return 0
end
function timer_event(ns)
for name, states in pairs(services) do
inject_payload('json', payload_name,
cjson.encode({vip_active_at=vip_active_at, name=name, states=states}))
end
end

70
deployment_scripts/puppet/modules/lma_collector/files/plugins/filters/service_status.lua
View File

@@ -14,18 +14,15 @@
require 'cjson'
require 'string'
require 'math'
local floor = math.floor
local max = math.max
local utils = require 'lma_utils'
_PRESERVATION_VERSION = 2
_PRESERVATION_VERSION = 3
-- variables with global scope are preserved between restarts
all_service_status = {}
-- local scope variables
local timeout = read_config("timeout") or 60
local hostname
local datapoints = {}
local timeout = (read_config("timeout") or 60) * 1e9
function process_message ()
local ok, data = pcall(cjson.decode, read_message("Payload"))
@@ -33,8 +30,7 @@ function process_message ()
return -1
end
local timestamp = read_message('Timestamp')
local ts = floor(timestamp/1e6) -- in ms
hostname = read_message("Hostname")
local hostname = read_message("Hostname")
local service_name = data.name
local states = data.states
@@ -49,23 +45,25 @@ function process_message ()
if not all_service_status[service_name] then all_service_status[service_name] = {} end
if states.workers then
worker_status = compute_status(events, not_up_status, ts, 'workers', service_name, states.workers, true)
worker_status = compute_status(events, not_up_status, timestamp, 'workers', service_name, states.workers, true)
end
if states.check_api then
check_api_status = compute_status(events, not_up_status, ts, 'check_api', service_name, states.check_api, false)
check_api_status = compute_status(events, not_up_status, timestamp, 'check_api', service_name, states.check_api, false)
end
if states.haproxy then
haproxy_server_status = compute_status(events, not_up_status, ts, 'haproxy', service_name, states.haproxy, true)
haproxy_server_status = compute_status(events, not_up_status, timestamp, 'haproxy', service_name, states.haproxy, true)
end
global_status = max(worker_status, check_api_status, haproxy_server_status)
-- global service status
utils.add_metric(datapoints,
string.format('%s.openstack.%s.status', hostname, service_name),
{ts, global_status})
utils.add_to_bulk_metric(
string.format('openstack_%s_status', service_name),
global_status
)
utils.inject_bulk_metric(timestamp, hostname, 'service_status_filter')
-- only emit status if the public vip is active
if not expired(ts, data.vip_active_at) then
if not expired(data.vip_active_at) then
local prev = all_service_status[service_name].global_status or utils.global_status_map.UNKNOWN
local updated
updated = (prev ~= global_status or #events > 0)
@@ -76,17 +74,13 @@ function process_message ()
if #events > 0 then
details = cjson.encode(events)
end
utils.inject_status_message(timestamp, service_name,
global_status, prev,
updated, details)
utils.inject_status_message(
timestamp, service_name, global_status, prev, updated, details
)
end
all_service_status[service_name].global_status = global_status
if #datapoints > 0 then
inject_payload("json", "influxdb", cjson.encode(datapoints))
datapoints = {}
end
return 0
end
@@ -183,10 +177,13 @@ function compute_status(events, not_up_status, current_time, elts_name, name, st
utils.service_status_to_label_map[DOWN],
event_detail)
end
utils.add_metric(datapoints, string.format('%s.openstack.%s.%s.%s.status',
hostname, name, worker.group_name, worker_name),
{current_time, utils.service_status_map.DOWN})
utils.add_to_bulk_metric(
string.format('openstack_%s_%s_status', name, worker.group_name),
utils.service_status_map.DOWN,
{ service = worker_name}
)
end
-- elements down or degraded
for worker_name, worker in pairs(down_elts) do
local prev = get_previous_status(name, elts_name, worker_name)
@@ -198,11 +195,11 @@ function compute_status(events, not_up_status, current_time, elts_name, name, st
new_status = utils.service_status_map.DOWN
end
set_status(name, elts_name, worker_name, new_status)
utils.add_metric(datapoints,
string.format("%s.openstack.%s.%s.%s.status",
hostname, name, worker.group_name, worker_name),
{current_time, new_status})
utils.add_to_bulk_metric(
string.format('openstack_%s_%s_status', name, worker.group_name),
new_status,
{ service = worker_name}
)
if display_num then
event_detail = string.format("(%s/%s UP)", up_elements[worker_name],
total_elements[worker_name])
@@ -235,17 +232,16 @@ function compute_status(events, not_up_status, current_time, elts_name, name, st
utils.service_status_to_label_map[prev],
utils.service_status_to_label_map[UP])
end
utils.add_metric(datapoints, string.format("%s.openstack.%s.%s.%s.status",
hostname, name, worker.group_name, worker_name),
{current_time, utils.service_status_map.UP})
utils.add_to_bulk_metric(
string.format('openstack_%s_%s_status', name, worker.group_name),
utils.service_status_map.UP,
{ service = worker_name}
)
end
end
return service_status
end
function expired(current_time, last_time)
if last_time > 0 and current_time - last_time <= timeout * 1000 then
return false
end
return true
function expired(last_time)
return not (last_time > 0 and (read_message('Timestamp') - last_time) <= timeout)
end

251
deployment_scripts/puppet/modules/lma_collector/files/plugins/filters/service_status_legacy.lua Normal file
View File

@@ -0,0 +1,251 @@
-- Copyright 2015 Mirantis, Inc.
--
-- Licensed under the Apache License, Version 2.0 (the "License");
-- you may not use this file except in compliance with the License.
-- You may obtain a copy of the License at
--
-- http://www.apache.org/licenses/LICENSE-2.0
--
-- Unless required by applicable law or agreed to in writing, software
-- distributed under the License is distributed on an "AS IS" BASIS,
-- WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-- See the License for the specific language governing permissions and
-- limitations under the License.
require 'cjson'
require 'string'
require 'math'
local floor = math.floor
local max = math.max
local utils = require 'lma_utils'
_PRESERVATION_VERSION = 2
-- variables with global scope are preserved between restarts
all_service_status = {}
-- local scope variables
local timeout = read_config("timeout") or 60
local hostname
local datapoints = {}
function process_message ()
local ok, data = pcall(cjson.decode, read_message("Payload"))
if not ok then
return -1
end
local timestamp = read_message('Timestamp')
local ts = floor(timestamp/1e6) -- in ms
hostname = read_message("Hostname")
local service_name = data.name
local states = data.states
local worker_status = -1
local check_api_status = -1
local haproxy_server_status = -1
local global_status
local events = {}
local not_up_status = {}
local msg_event
if not all_service_status[service_name] then all_service_status[service_name] = {} end
if states.workers then
worker_status = compute_status(events, not_up_status, ts, 'workers', service_name, states.workers, true)
end
if states.check_api then
check_api_status = compute_status(events, not_up_status, ts, 'check_api', service_name, states.check_api, false)
end
if states.haproxy then
haproxy_server_status = compute_status(events, not_up_status, ts, 'haproxy', service_name, states.haproxy, true)
end
global_status = max(worker_status, check_api_status, haproxy_server_status)
-- global service status
utils.add_metric(datapoints,
string.format('%s.openstack.%s.status', hostname, service_name),
{ts, global_status})
-- only emit status if the public vip is active
if not expired(ts, data.vip_active_at) then
local prev = all_service_status[service_name].global_status or utils.global_status_map.UNKNOWN
local updated
updated = (prev ~= global_status or #events > 0)
-- always append not UP status elements in details
for k, v in pairs(not_up_status) do events[#events+1] = v end
local details = ''
if #events > 0 then
details = cjson.encode(events)
end
utils.inject_status_message(timestamp, service_name,
global_status, prev,
updated, details)
end
all_service_status[service_name].global_status = global_status
if #datapoints > 0 then
inject_payload("json", "influxdb", cjson.encode(datapoints))
datapoints = {}
end
return 0
end
function get_previous_status(service_name, top_entry, name)
if not all_service_status[service_name] then
all_service_status[service_name] = {}
end
if not all_service_status[service_name][top_entry] then
all_service_status[service_name][top_entry] = {}
end
if not all_service_status[service_name][top_entry][name] then
all_service_status[service_name][top_entry][name] = utils.service_status_map.UNKNOWN
end
return all_service_status[service_name][top_entry][name]
end
function set_status(service_name, top_entry, name, status)
all_service_status[service_name][top_entry][name] = status
end
function compute_status(events, not_up_status, current_time, elts_name, name, states, display_num)
local down_elts = {}
local down_elts_count = 0
local zero_up = {}
local zero_up_count = 0
local one_up = {}
local one_disabled = {}
local one_disabled_count = 0
local service_status = utils.service_status_map.UNKNOWN
local up_elements = {}
local total_elements = {}
for worker, worker_data in pairs(states) do
if not total_elements[worker] then
total_elements[worker] = 0
end
if not up_elements[worker] then
up_elements[worker] = 0
end
for state, data in pairs(worker_data) do
if not expired(current_time, data.last_seen) then
total_elements[worker] = total_elements[worker] + data.value
if state == utils.state_map.DOWN and data.value > 0 then
down_elts[worker] = data
down_elts_count = down_elts_count + 1
end
if state == utils.state_map.UP then
if data.value > 0 then
one_up[worker] = data
else
zero_up[worker] = data
zero_up_count = zero_up_count + 1
end
up_elements[worker] = data.value
end
if state == utils.state_map.DISABLED and data.value > 0 then
one_disabled[worker] = data
one_disabled_count = one_disabled_count + 1
end
end
end
end
-- general element status
if zero_up_count > 0 then
service_status = utils.service_status_map.DOWN
elseif down_elts_count > 0 then
service_status = utils.service_status_map.DEGRADED
elseif down_elts_count == 0 then
service_status = utils.service_status_map.UP
end
-- elements clearly down
for worker_name, worker in pairs(zero_up) do
local prev = get_previous_status(name, elts_name, worker_name)
local DOWN = utils.service_status_map.DOWN
local event_detail = ""
set_status(name, elts_name, worker_name, DOWN)
if display_num then
event_detail = string.format("(%s/%s UP)", up_elements[worker_name],
total_elements[worker_name])
end
if prev and prev ~= DOWN then
events[#events+1] = string.format("%s %s %s -> %s %s", worker_name,
worker.group_name,
utils.service_status_to_label_map[prev],
utils.service_status_to_label_map[DOWN],
event_detail)
else
not_up_status[#not_up_status+1] = string.format("%s %s %s %s",
worker_name,
worker.group_name,
utils.service_status_to_label_map[DOWN],
event_detail)
end
utils.add_metric(datapoints, string.format('%s.openstack.%s.%s.%s.status',
hostname, name, worker.group_name, worker_name),
{current_time, utils.service_status_map.DOWN})
end
-- elements down or degraded
for worker_name, worker in pairs(down_elts) do
local prev = get_previous_status(name, elts_name, worker_name)
local new_status
local event_detail
if one_up[worker_name] then
new_status = utils.service_status_map.DEGRADED
else
new_status = utils.service_status_map.DOWN
end
set_status(name, elts_name, worker_name, new_status)
utils.add_metric(datapoints,
string.format("%s.openstack.%s.%s.%s.status",
hostname, name, worker.group_name, worker_name),
{current_time, new_status})
if display_num then
event_detail = string.format("(%s/%s UP)", up_elements[worker_name],
total_elements[worker_name])
else
event_detail = ""
end
if prev ~= new_status then
events[#events+1] = string.format("%s %s %s -> %s %s", worker_name,
worker.group_name,
utils.service_status_to_label_map[prev],
utils.service_status_to_label_map[new_status],
event_detail)
elseif not zero_up[worker_name] then
not_up_status[#not_up_status+1] = string.format("%s %s %s %s", worker_name,
worker.group_name,
utils.service_status_to_label_map[new_status],
event_detail)
end
end
-- elements up
for worker_name, worker in pairs(one_up) do
if not zero_up[worker_name] and not down_elts[worker_name] then
local prev = get_previous_status(name, elts_name, worker_name)
local UP = utils.service_status_map.UP
set_status(name, elts_name, worker_name, UP)
if prev and prev ~= utils.service_status_map.UP then
events[#events+1] = string.format("%s %s %s -> %s", worker_name,
worker.group_name,
utils.service_status_to_label_map[prev],
utils.service_status_to_label_map[UP])
end
utils.add_metric(datapoints, string.format("%s.openstack.%s.%s.%s.status",
hostname, name, worker.group_name, worker_name),
{current_time, utils.service_status_map.UP})
end
end
return service_status
end
function expired(current_time, last_time)
if last_time > 0 and current_time - last_time <= timeout * 1000 then
return false
end
return true
end

24
deployment_scripts/puppet/modules/lma_collector/manifests/metrics/service_status.pp
View File

@@ -13,26 +13,26 @@
# under the License.
#
class lma_collector::metrics::service_status (
$metrics_regexp = $lma_collector::params::service_status_metrics_regexp,
$payload_name = $lma_collector::params::service_status_payload_name,
$metrics_matcher = $lma_collector::params::service_status_metrics_matcher,
$timeout = $lma_collector::params::service_status_timeout,
){
include heka::params
) inherits lma_collector::params {
validate_array($metrics_regexp)
validate_string($metrics_regexp)
$payload_name = $lma_collector::params::service_status_payload_name
if (size(metrics_regexp) > 0){
heka::filter::sandbox { 'service_accumulator_states':
config_dir => $lma_collector::params::config_dir,
filename => "${lma_collector::params::plugins_dir}/filters/service_accumulator_states.lua",
message_matcher => inline_template('<%= @metrics_regexp.collect{|x| "Fields[name] =~ /%s/" % x}.join(" || ") %>'),
ticker_interval => $lma_collector::params::service_status_interval,
preserve_data => true,
config => {
config_dir => $lma_collector::params::config_dir,
filename => "${lma_collector::params::plugins_dir}/filters/service_accumulator_states.lua",
message_matcher => $metrics_matcher,
ticker_interval => $lma_collector::params::service_status_interval,
preserve_data => true,
config => {
inject_payload_name => $payload_name,
},
notify => Class['lma_collector::service'],
notify => Class['lma_collector::service'],
}
heka::filter::sandbox { 'service_status':

49
deployment_scripts/puppet/modules/lma_collector/manifests/metrics/service_status_legacy.pp Normal file
View File

@@ -0,0 +1,49 @@
# Copyright 2015 Mirantis, Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
#
class lma_collector::metrics::service_status_legacy (
$metrics_regexp = $lma_collector::params::service_status_metrics_regexp_legacy,
$payload_name = $lma_collector::params::service_status_payload_name,
$timeout = $lma_collector::params::service_status_timeout,
){
include heka::params
validate_array($metrics_regexp)
if (size(metrics_regexp) > 0){
heka::filter::sandbox { 'service_accumulator_states':
config_dir => $lma_collector::params::config_dir,
filename => "${lma_collector::params::plugins_dir}/filters/service_accumulator_states_legacy.lua",
message_matcher => inline_template('<%= @metrics_regexp.collect{|x| "Fields[name] =~ /%s/" % x}.join(" || ") %>'),
ticker_interval => $lma_collector::params::service_status_interval,
preserve_data => true,
config => {
inject_payload_name => $payload_name,
},
notify => Class['lma_collector::service'],
}
heka::filter::sandbox { 'service_status':
config_dir => $lma_collector::params::config_dir,
filename => "${lma_collector::params::plugins_dir}/filters/service_status_legacy.lua",
message_matcher => "Fields[payload_type] == 'json' && Fields[payload_name] == '${payload_name}'",
preserve_data => true,
config => {
timeout => $timeout,
},
notify => Class['lma_collector::service'],
}
}
}

28
deployment_scripts/puppet/modules/lma_collector/manifests/params.pp
View File

@@ -105,15 +105,25 @@ class lma_collector::params {
$service_status_payload_name = 'service_status'
# Catch all metrics used to compute OpenStack service statutes
$service_status_metrics_regexp = [
'^openstack.(nova|cinder|neutron).(services|agents).*(up|down|disabled)$',
# Exception for mysqld backend because the MySQL service status is
# computed by a dedicated filter and this avoids to send an annoying
# status Heka message.
'^haproxy.backend.(horizon|nova|cinder|neutron|ceilometer|keystone|swift|heat|glance|radosgw)(-.+)?.servers.(down|up)$',
'^pacemaker.resource.vip__public.active$',
'^openstack.*check_api$'
]
$service_status_metrics_regexp_legacy = [
'^openstack.(nova|cinder|neutron).(services|agents).*(up|down|disabled)$',
# Exception for mysqld backend because the MySQL service status is
# computed by a dedicated filter and this avoids to send an annoying
# status Heka message.
'^haproxy.backend.(horizon|nova|cinder|neutron|ceilometer|keystone|swift|heat|glance|radosgw)(-.+)?.servers.(down|up)$',
'^pacemaker.resource.vip__public.active$',
'^openstack.*check_api$'
]
$service_status_metrics_matcher = join([
'(Type == \'metric\' || Type == \'heka.sandbox.metric\') && ',
'(Fields[name] =~ /^openstack_(nova|cinder|neutron)_(services|agents)$/ || ',
# Exception for mysqld backend because the MySQL service status is
# computed by a dedicated filter and this avoids to send an annoying
# status Heka message.
'(Fields[name] == \'haproxy_backend_servers\' && Fields[backend] !~ /mysql/) || ',
'(Fields[name] == \'pacemaker_local_resource_active\' && Fields[resource] == \'vip__public\') || ',
'Fields[name] =~ /^openstack.*check_api$/)'
], '')
$worker_report_interval = 60
$worker_downtime_factor = 2