class Sensu::Server::Process
Constants
- DEFAULT_HANDLER_NAME
- EVENT_FLAPPING_ACTION
- METRIC_CHECK_TYPE
- STANDARD_CHECK_TYPE
- TASKS
Attributes
Public Class Methods
Override Daemon
initialize() to support Sensu
server tasks and the handling event count.
@param options [Hash]
Sensu::Daemon::new
# File lib/sensu/server/process.rb, line 44 def initialize(options={}) super @tasks = [] @timers[:tasks] = {} TASKS.each do |task| @timers[:tasks][task.to_sym] = [] end @in_progress = Hash.new(0) end
Create an instance of the Sensu
server process, start the server within the EventMachine event loop, and set up server process signal traps (for stopping).
@param options [Hash]
# File lib/sensu/server/process.rb, line 32 def self.run(options={}) server = self.new(options) EM::run do server.start server.setup_signal_traps end end
Public Instance Methods
Add a check result to one or more aggregates. The aggregate name is determined by the value of check `:aggregates` array, if present, and falling back to `:aggregate` otherwise.
When one or more aggregates are specified as `:aggregates`, the client name and check are updated on each aggregate.
When no aggregates are specified as `:aggregates`, and `:aggregate` is `true` (legacy), the check `:name` is used as the aggregate name.
When no aggregates are specified as `:aggregates` and check `:aggregate` is a string, it used as the aggregate name.
This method will add the client name to configured aggregates, all other processing (e.g. counters) is done by the Sensu
API
on request.
@param client [Hash] @param check [Hash]
# File lib/sensu/server/process.rb, line 341 def aggregate_check_result(client, check) check_aggregate = (check[:aggregate].is_a?(String) ? check[:aggregate] : check[:name]) aggregate_list = Array(check[:aggregates] || check_aggregate) aggregate_list.each do |aggregate| @logger.debug("adding check result to aggregate", { :aggregate => aggregate, :client => client, :check => check }) aggregate_member = "#{client[:name]}:#{check[:name]}" @redis.sadd("aggregates:#{aggregate}", aggregate_member) do @redis.sadd("aggregates", aggregate) end end end
Bootstrap the Sensu
server process, setting up the keepalive and check result consumers, and attemping to carry out Sensu
server tasks. This method sets the process/daemon `@state` to `:running`.
# File lib/sensu/server/process.rb, line 1563 def bootstrap setup_keepalives setup_results setup_task_elections setup_server_registry_updater setup_tessen @state = :running end
Calculate a check request splay, taking into account the current time and the request interval to ensure it's consistent between process restarts.
@param check [Hash] definition.
# File lib/sensu/server/process.rb, line 1017 def calculate_check_request_splay(check) splay_hash = Digest::MD5.digest(check[:name]).unpack('Q<').first current_time = (Time.now.to_f * 1000).to_i (splay_hash - current_time) % (check[:interval] * 1000) / 1000.0 end
Determine if a check state is flapping, rapidly changing between an OK and non-OK state. Flap detection is only done for checks that have defined low and hight flap detection thresholds, `:low_flap_threshold` and `:high_flap_threshold`. The `check_history()` method provides the check history and more importantly the total state change precentage value that is compared with the configured thresholds defined in the check data. If a check hasn't been flapping, the `:total_state_change` must be equal to or higher than the `:high_flap_threshold` to be changed to flapping. If a check has been flapping, the `:total_state_change` must be equal to or lower than the `:low_flap_threshold` to no longer be flapping. This method uses the same algorithm as Nagios: nagios.sourceforge.net/docs/3_0/flapping.html
@param stored_event [Hash] @param check [Hash] @return [TrueClass, FalseClass]
# File lib/sensu/server/process.rb, line 479 def check_flapping?(stored_event, check) if check.has_key?(:low_flap_threshold) && check.has_key?(:high_flap_threshold) if check[:low_flap_threshold].is_a?(Integer) && check[:high_flap_threshold].is_a?(Integer) was_flapping = stored_event && stored_event[:action] == EVENT_FLAPPING_ACTION if was_flapping check[:total_state_change] > check[:low_flap_threshold] else check[:total_state_change] >= check[:high_flap_threshold] end else details = {:check => check} details[:client] = stored_event[:client] if stored_event @logger.error("invalid check flap thresholds", details) false end else false end end
Fetch the execution history for a client/check pair, the 21 most recent check result statuses. This method also calculates the total state change percentage for the history, this value is use for check state flap detection, using a similar algorithm to Nagios: nagios.sourceforge.net/docs/3_0/flapping.html
@param client [Hash] @param check [Hash] @yield [history, total_state_change] callback/block to call
with the check history and calculated total state change value.
@yieldparam history [Array] containing the last 21 check
result exit status codes.
@yieldparam total_state_change [Float] percentage for the
check history (exit status codes).
@yieldparam last_ok [Integer] execution timestamp of the last
OK check result.
# File lib/sensu/server/process.rb, line 437 def check_history(client, check) history_key = "history:#{client[:name]}:#{check[:name]}" @redis.lrange(history_key, -21, -1) do |history| total_state_change = 0 unless history.length < 21 state_changes = 0 change_weight = 0.8 previous_status = history.first history.each do |status| unless status == previous_status state_changes += change_weight end change_weight += 0.02 previous_status = status end total_state_change = (state_changes.fdiv(20) * 100).to_i end @redis.get("#{history_key}:last_ok") do |last_ok| last_ok = last_ok.to_i unless last_ok.nil? yield(history, total_state_change, last_ok) end end end
Complete in progress work and then call the provided callback. This method will wait until all counters stored in the `@in_progress` hash equal `0`.
@yield [] callback/block to call when in progress work is
completed.
# File lib/sensu/server/process.rb, line 1549 def complete_in_progress @logger.info("completing work in progress", :in_progress => @in_progress) retry_until_true do if @in_progress.values.all? { |count| count == 0 } yield true end end end
Create a check request proc, used to publish check requests to for a check to the Sensu
transport. Check requests are not published if subdued. This method determines if a check uses proxy check requests and calls the appropriate check request publish method.
@param check [Hash] definition.
# File lib/sensu/server/process.rb, line 979 def create_check_request_proc(check) Proc.new do unless check_subdued?(check) if check[:proxy_requests] create_proxy_check_requests(check) else publish_check_request(check) end else @logger.info("check request was subdued", :check => check) end end end
Create check TTL results. This method will retrieve check results from the registry and determine the time since their last check execution (in seconds). If the time since last execution is equal to or greater than the defined check TTL, a warning check result is published with the appropriate check output.
@param ttl_keys [Array] of TTL keys. @param interval [Integer] to use for the check TTL result
interval.
@yield [] callback/block called after the check TTL results
have been created.
# File lib/sensu/server/process.rb, line 1223 def create_check_ttl_results(ttl_keys, interval=30) result_keys = ttl_keys.map { |ttl_key| "result:#{ttl_key}" } @redis.mget(*result_keys) do |result_json_objects| result_json_objects.each_with_index do |result_json, index| unless result_json.nil? check = Sensu::JSON.load(result_json) next unless check[:ttl] && check[:executed] && !check[:force_resolve] time_since_last_execution = Time.now.to_i - check[:executed] if time_since_last_execution >= check[:ttl] client_name = ttl_keys[index].split(":").first keepalive_event_exists?(client_name) do |event_exists| unless event_exists check[:output] = "Last check execution was " check[:output] << "#{time_since_last_execution} seconds ago" check[:status] = check[:ttl_status] || 1 check[:interval] = interval publish_check_result(client_name, check) end end end else @redis.srem("ttl", ttl_keys[index]) end end yield end end
Create a blank client (data). Only the client name is known, the other client attributes must be updated via the API
(POST /clients:client). Dynamically created clients and those updated via the API
will have client keepalives disabled by default, `:keepalives` is set to `false`.
@param name [String] to use for the client. @return [Hash] client.
# File lib/sensu/server/process.rb, line 640 def create_client(name) { :name => name, :address => "unknown", :subscriptions => ["client:#{name}"], :keepalives => false, :version => VERSION, :timestamp => Time.now.to_i } end
Create client keepalive check results. This method will retrieve clients from the registry, creating a keepalive check definition for each client, using the `create_keepalive_check()` method, containing client specific keepalive thresholds. If the time since the latest keepalive is equal to or greater than a threshold, the check `:output` is set to a descriptive message, and `:status` is set to the appropriate non-zero value. If a client has been sending keepalives, `:output` and `:status` are set to indicate an OK state. The `publish_check_result()` method is used to publish the client keepalive check results.
@param clients [Array] of client names. @yield [] callback/block called after the client keepalive
check results have been created.
# File lib/sensu/server/process.rb, line 1143 def create_client_keepalive_check_results(clients) client_keys = clients.map { |client_name| "client:#{client_name}" } @redis.mget(*client_keys) do |client_json_objects| client_json_objects.each do |client_json| unless client_json.nil? client = Sensu::JSON.load(client_json) next if client[:keepalives] == false check, unmatched_tokens = create_keepalive_check(client) time_since_last_keepalive = Time.now.to_i - client[:timestamp] check[:output] = "No keepalive sent from client for " check[:output] << "#{time_since_last_keepalive} seconds" case when time_since_last_keepalive >= check[:thresholds][:critical] check[:output] << " (>=#{check[:thresholds][:critical]})" check[:status] = 2 when time_since_last_keepalive >= check[:thresholds][:warning] check[:output] << " (>=#{check[:thresholds][:warning]})" check[:status] = 1 else check[:output] = "Keepalive sent from client " check[:output] << "#{time_since_last_keepalive} seconds ago" check[:status] = 0 end unless unmatched_tokens.empty? check[:output] << " - Unmatched client token(s): " + unmatched_tokens.join(", ") check[:status] = 1 if check[:status] == 0 end publish_check_result(client[:name], check) end end yield end end
Create and process a client registration event. A registration event is created when a Sensu
client is first added to the client registry. The `create_registration_check()` method is called to create a registration check definition for the client.
@param client [Hash] definition.
# File lib/sensu/server/process.rb, line 105 def create_client_registration_event(client) check = create_registration_check(client) create_event(client, check) do |event| event_bridges(event) process_event(event) end end
Create an event, using the provided client and check result data. Existing event data for the client/check pair is fetched from the event registry to be used in the composition of the new event. The silenced registry is used to determine if the event has been silenced.
@param client [Hash] @param check [Hash] @yield callback [event] callback/block called with the
resulting event.
@yieldparam event [Hash]
# File lib/sensu/server/process.rb, line 588 def create_event(client, check) check_history(client, check) do |history, total_state_change, last_ok| check[:history] = history check[:total_state_change] = total_state_change @redis.hget("events:#{client[:name]}", check[:name]) do |event_json| stored_event = event_json ? Sensu::JSON.load(event_json) : nil flapping = check_flapping?(stored_event, check) event = { :id => random_uuid, :client => client, :check => check, :occurrences => 1, :occurrences_watermark => 1, :last_ok => last_ok, :action => (flapping ? :flapping : :create), :timestamp => Time.now.to_i } if stored_event event[:id] = stored_event[:id] event[:last_state_change] = stored_event[:last_state_change] event[:occurrences] = stored_event[:occurrences] event[:occurrences_watermark] = stored_event[:occurrences_watermark] || event[:occurrences] end if check[:status] != 0 || flapping if history[-1] == history[-2] event[:occurrences] += 1 if event[:occurrences] > event[:occurrences_watermark] event[:occurrences_watermark] = event[:occurrences] end else event[:occurrences] = 1 event[:last_state_change] = event[:timestamp] end elsif stored_event event[:last_state_change] = event[:timestamp] event[:action] = :resolve end event_silenced?(event) do |event| yield(event) end end end end
Create a keepalive check definition for a client. Client
definitions may contain `:keepalive` configuration, containing specific thresholds and handler information. The keepalive check definition creation begins with default thresholds, and sets the `:handler` to `keepalive`, if the handler has a local definition. If the client provides its own `:keepalive` configuration, it's deep merged with the defaults. The check `:name`, `:issued`, and `:executed` values are always overridden to guard against an invalid definition.
@return [Array] check definition, unmatched client tokens
# File lib/sensu/server/process.rb, line 1107 def create_keepalive_check(client) check = { :thresholds => { :warning => 120, :critical => 180 } } if @settings.handler_exists?(:keepalive) check[:handler] = "keepalive" end if @settings[:sensu][:keepalives] check = deep_merge(check, @settings[:sensu][:keepalives]) end if client.has_key?(:keepalive) check = deep_merge(check, client[:keepalive]) end timestamp = Time.now.to_i check.merge!(:name => "keepalive", :issued => timestamp, :executed => timestamp) object_substitute_tokens(check, client) end
Create a lock timestamp (integer), current time including milliseconds. This method is used by Sensu
server task election.
@return [Integer]
# File lib/sensu/server/process.rb, line 1299 def create_lock_timestamp (Time.now.to_f * 10000000).to_i end
Create and publish one or more proxy check requests. This method iterates through the Sensu
client registry for clients that matched provided proxy request client attributes. A proxy check request is created for each client in the registry that matches the proxy request client attributes. Proxy check requests have their client tokens subsituted by the associated client attributes values. The `determine_matching_clients()` method is used to fetch and inspect each slide of clients from the registry, returning those that match the configured proxy request client attributes. A relatively small clients slice size (20) is used to reduce the number of clients inspected within a single tick of the EM reactor. The `publish_proxy_check_requests()` method is used to iterate through the matching Sensu
clients, creating their own unique proxy check request, substituting client tokens, and then publishing them to the targetted subscriptions.
@param check [Hash] definition.
# File lib/sensu/server/process.rb, line 951 def create_proxy_check_requests(check) client_attributes = check[:proxy_requests][:client_attributes] unless client_attributes.empty? @redis.smembers("clients") do |clients| client_count = clients.length proxy_check_requests = Proc.new do |matching_clients, slice_start, slice_size| unless slice_start > client_count - 1 clients_slice = clients.slice(slice_start..slice_size) determine_matching_clients(clients_slice, client_attributes) do |additional_clients| matching_clients += additional_clients proxy_check_requests.call(matching_clients, slice_start + 20, slice_size + 20) end else publish_proxy_check_requests(matching_clients, check) end end proxy_check_requests.call([], 0, 19) end end end
Create a registration check definition for a client. Client
definitions may contain `:registration` configuration, containing custom attributes and handler information. By default, the registration check definition sets the `:handler` to `registration`. If the client provides its own `:registration` configuration, it's deep merged with the defaults. The check `:name`, `:output`, `:issued`, and `:executed` values are always overridden to guard against an invalid definition.
# File lib/sensu/server/process.rb, line 83 def create_registration_check(client) check = {:handler => "registration", :status => 1} if client.has_key?(:registration) check = deep_merge(check, client[:registration]) end timestamp = Time.now.to_i overrides = { :name => "registration", :output => "new client registration", :issued => timestamp, :executed => timestamp } check.merge(overrides) end
Derive an array of handler definitions from a list of handler names. This method first checks for the existence of standard handlers, followed by handler extensions. If a handler does not exist for a name, it is logged and ignored. Duplicate handler definitions are removed.
@param handler_list [Array] @param depth [Integer] of handler set expansion. @return [Array]
# File lib/sensu/server/process.rb, line 268 def derive_handlers(handler_list, depth=0) handler_list.compact.map { |handler_name| case when @settings.handler_exists?(handler_name) handler = @settings[:handlers][handler_name].merge(:name => handler_name) expand_handler_sets(handler, depth) when @extensions.handler_exists?(handler_name) @extensions[:handlers][handler_name] else @logger.error("unknown handler", :handler_name => handler_name) nil end }.flatten.compact.uniq end
Determine and return clients from the registry that match a set of attributes.
@param clients [Array] of client names. @param attributes [Hash] @yield [Array] callback/block called after determining the
matching clients, returning them as a block parameter.
# File lib/sensu/server/process.rb, line 852 def determine_matching_clients(clients, attributes) client_keys = clients.map { |client_name| "client:#{client_name}" } @redis.mget(*client_keys) do |client_json_objects| matching_clients = [] client_json_objects.each do |client_json| unless client_json.nil? client = Sensu::JSON.load(client_json) if attributes_match?(client, attributes) matching_clients << client end end end yield(matching_clients) end end
Determine stale check results, those that have not executed in a specified amount of time (check TTL). This method iterates through stored check results that have a defined TTL value (in seconds). The `create_check_ttl_results()` method is used to inspect each check result, calculating their time since last check execution (in seconds). If the time since last execution is equal to or greater than the check TTL, a warning check result is published with the appropriate check output. A relatively small check results slice size (20) is used to reduce the number of check results inspected within a single tick of the EM reactor.
@param interval [Integer] to use for the check TTL result
interval.
# File lib/sensu/server/process.rb, line 1265 def determine_stale_check_results(interval=30) @logger.info("determining stale check results (ttl)") @redis.smembers("ttl") do |ttl_keys| ttl_key_count = ttl_keys.length ttl_check_results = Proc.new do |slice_start, slice_size| unless slice_start > ttl_key_count - 1 ttl_keys_slice = ttl_keys.slice(slice_start..slice_size) create_check_ttl_results(ttl_keys_slice, interval) do ttl_check_results.call(slice_start + 20, slice_size + 20) end end end ttl_check_results.call(0, 19) end end
Determine stale clients, those that have not sent a keepalive in a specified amount of time. This method iterates through the client registry, creating a keepalive check result for each client. The `create_client_keepalive_check_results()` method is used to inspect and create keepalive check results for each slice of clients from the registry. A relatively small clients slice size (20) is used to reduce the number of clients inspected within a single tick of the EM reactor.
# File lib/sensu/server/process.rb, line 1185 def determine_stale_clients @logger.info("determining stale clients") @redis.smembers("clients") do |clients| client_count = clients.length keepalive_check_results = Proc.new do |slice_start, slice_size| unless slice_start > client_count - 1 clients_slice = clients.slice(slice_start..slice_size) create_client_keepalive_check_results(clients_slice) do keepalive_check_results.call(slice_start + 20, slice_size + 20) end end end keepalive_check_results.call(0, 19) end end
Run event bridge extensions, within the Sensu
EventMachine reactor (event loop). The extension API
`safe_run()` method is used to guard against most errors. Bridges are for relaying Sensu
event data to other services.
@param event [Hash]
# File lib/sensu/server/process.rb, line 312 def event_bridges(event) @extensions[:bridges].each do |name, bridge| bridge.safe_run(event) do |output, status| @logger.debug("bridge extension output", { :extension => bridge.definition, :output => output }) end end end
Determine if an event has been silenced. This method compiles an array of possible silenced registry entry keys for the event. An attempt is made to fetch one or more of the silenced registry entries to determine if the event has been silenced. The event data is updated to indicate if the event has been silenced. If the event is silenced and the event action is `:resolve`, silenced registry entries with `:expire_on_resolve` set to true will be deleted. Silencing is disabled for events with a check status of `0` (OK), unless the event action is `:resolve` or `:flapping`.
@param event [Hash] @yield callback [event] callback/block called after the event
data has been updated to indicate if it has been silenced.
# File lib/sensu/server/process.rb, line 513 def event_silenced?(event) event[:silenced] = false event[:silenced_by] = [] if event[:check][:status] != 0 || event[:action] != :create check_name = event[:check][:name] silenced_keys = event[:client][:subscriptions].map { |subscription| ["silence:#{subscription}:*", "silence:#{subscription}:#{check_name}"] }.flatten silenced_keys << "silence:*:#{check_name}" @redis.mget(*silenced_keys) do |silenced| silenced.compact! silenced.each do |silenced_json| silenced_info = Sensu::JSON.load(silenced_json) if silenced_info[:expire_on_resolve] && event[:action] == :resolve silenced_key = "silence:#{silenced_info[:id]}" @redis.srem("silenced", silenced_key) @redis.del(silenced_key) elsif silenced_info[:begin].nil? || silenced_info[:begin] <= Time.now.to_i event[:silenced_by] << silenced_info[:id] end end event[:silenced] = !event[:silenced_by].empty? yield(event) end else yield(event) end end
Expand event handler sets, creating an array of handler definitions. Handler sets cannot be deeply nested (by choice), this method will return `nil` if an attempt is made to deeply nest. If the provided handler definition is not a set, it is returned.
@param handler [Hash] definition. @param depth [Integer] of the expansion. @return [Array, Hash, Nil]
# File lib/sensu/server/process.rb, line 246 def expand_handler_sets(handler, depth=0) if handler[:type] == "set" if depth < 2 derive_handlers(handler[:handlers], depth + 1) else @logger.error("handler sets cannot be deeply nested", :handler => handler) nil end else handler end end
Determine if a keepalive event exists for a client.
@param client_name [String] name of client to look up in event registry. @return [TrueClass, FalseClass]
# File lib/sensu/server/process.rb, line 708 def keepalive_event_exists?(client_name) @redis.hexists("events:#{client_name}", "keepalive") do |event_exists| yield(event_exists) end end
Determine if a transport message is under the optional configured max message size. This method helps prevent oversized messages from consuming memory and being persisted to the datastore.
@param message [String] @return [TrueClass,FalseClass]
# File lib/sensu/server/process.rb, line 185 def message_size_ok?(message) if @settings[:sensu][:server] && @settings[:sensu][:server][:max_message_size] message_size = message.bytesize max_message_size = @settings[:sensu][:server][:max_message_size] if message_size <= max_message_size true else @logger.error("message exceeds the configured max message size", { :max_message_size => max_message_size, :message_size => message_size, :message => message }) false end else true end end
Pause the Sensu
server process, unless it is being paused or has already been paused. The process/daemon `@state` is first set to `:pausing`, to indicate that it's in progress. All run timers are cancelled, their references are cleared, and Tessen
is stopped. The Sensu
server will unsubscribe from all transport subscriptions, relinquish any Sensu
server tasks, then set the process/daemon `@state` to `:paused`.
# File lib/sensu/server/process.rb, line 1589 def pause unless @state == :pausing || @state == :paused @state = :pausing @timers[:run].each do |timer| timer.cancel end @timers[:run].clear @tessen.stop if @tessen unsubscribe relinquish_tasks @state = :paused end end
Process
a check result, storing its data, inspecting its contents, and taking the appropriate actions (eg. update the event registry). The `@in_progress` counter is incremented by `1` prior to check result processing and then decremented by `1` after updating the event registry. A check result must have a valid client name, associated with a client in the registry or one will be created. If a local check definition exists for the check name, and the check result is not from a standalone check execution, it's merged with the check result for more context.
@param result [Hash] data.
# File lib/sensu/server/process.rb, line 726 def process_check_result(result) @in_progress[:check_results] += 1 @logger.debug("processing result", :result => result) retrieve_client(result) do |client| unless client.nil? check = case when @settings.check_exists?(result[:check][:name]) && !result[:check][:standalone] @settings[:checks][result[:check][:name]].merge(result[:check]) else result[:check] end check[:type] ||= STANDARD_CHECK_TYPE check[:origin] = result[:client] if check[:source] if @settings.check_exists?(check[:name]) && client[:type] == "proxy" check[:command] = @settings[:checks][check[:name].to_sym][:command] end aggregate_check_result(client, check) if check[:aggregates] || check[:aggregate] store_check_result(client, check) do create_event(client, check) do |event| event_bridges(event) update_event_registry(event) do |process| process_event(event) if process @in_progress[:check_results] -= 1 end end end else @logger.warn("halting result processing", :result => result) @in_progress[:check_results] -= 1 end end end
Process
an initial client registration, when it is first added to the client registry. If a registration handler is defined or the client specifies one, a client registration event is created and processed (handled, etc.) for the client (`create_client_registration_event()`).
@param client [Hash] definition.
# File lib/sensu/server/process.rb, line 120 def process_client_registration(client) if @settings.handler_exists?("registration") || client[:registration] create_client_registration_event(client) end end
Process
an event: filter -> mutate -> handle.
This method determines the appropriate handlers for an event, filtering and mutating the event data for each of them. The `@in_progress` counter is incremented by `1`, for each event handler chain (filter -> mutate -> handle).
@param event [Hash]
# File lib/sensu/server/process.rb, line 291 def process_event(event) log_level = event[:check][:type] == METRIC_CHECK_TYPE ? :debug : :info @logger.send(log_level, "processing event", :event => event) handler_list = Array((event[:check][:handlers] || event[:check][:handler]) || DEFAULT_HANDLER_NAME) handlers = derive_handlers(handler_list) handlers.each do |handler| @in_progress[:events] += 1 filter_event(handler, event) do |event| mutate_event(handler, event) do |event_data| handle_event(handler, event_data, event[:id]) end end end end
Publish a check request to the Transport. A check request is composed of a check definition (minus `:subscribers` and `:interval`) and an `:issued` timestamp. The check request is published to a Transport pipe, for each of the check `:subscribers` in its definition, eg. “webserver”. JSON serialization is used when publishing the check request payload to the Transport pipes. Transport errors are logged.
@param check [Hash] definition.
# File lib/sensu/server/process.rb, line 822 def publish_check_request(check) payload = check.reject do |key, value| [:subscribers, :interval].include?(key) end payload[:issued] = Time.now.to_i @logger.info("publishing check request", { :payload => payload, :subscribers => check[:subscribers] }) check[:subscribers].each do |subscription| options = transport_publish_options(subscription, Sensu::JSON.dump(payload)) @transport.publish(*options) do |info| if info[:error] @logger.error("failed to publish check request", { :subscription => subscription, :payload => payload, :error => info[:error].to_s }) end end end end
Publish a check result to the Transport for processing. A check result is composed of a client name and a check definition, containing check `:output` and `:status`. A client signature is added to the check result payload if one is registered for the client. JSON serialization is used when publishing the check result payload to the Transport pipe. Transport errors are logged.
@param client_name [String] @param check [Hash]
# File lib/sensu/server/process.rb, line 1077 def publish_check_result(client_name, check) payload = { :client => client_name, :check => check } @redis.get("client:#{client_name}:signature") do |signature| payload[:signature] = signature if signature @logger.debug("publishing check result", :payload => payload) @transport.publish(:direct, "results", Sensu::JSON.dump(payload)) do |info| if info[:error] @logger.error("failed to publish check result", { :payload => payload, :error => info[:error].to_s }) end end end end
Publish a proxy check request for a client. This method substitutes client tokens in the check definition prior to publish the check request. If there are unmatched client tokens, a warning is logged, and a check request is not published.
@param client [Hash] definition. @param check [Hash] definition.
# File lib/sensu/server/process.rb, line 876 def publish_proxy_check_request(client, check) @logger.debug("creating a proxy check request", { :client => client, :check => check }) proxy_check, unmatched_tokens = object_substitute_tokens(deep_dup(check.dup), client) if unmatched_tokens.empty? proxy_check[:source] ||= client[:name] publish_check_request(proxy_check) else @logger.warn("failed to publish a proxy check request", { :reason => "unmatched client tokens", :unmatched_tokens => unmatched_tokens, :client => client, :check => check }) end end
Publish proxy check requests for one or more clients. This method can optionally splay proxy check requests, evenly, over a period of time, determined by the check interval and a configurable splay coverage percentage. For example, splay proxy check requests over 60s * 90%, 54s, leaving 6s for the last proxy check execution before the the next round of proxy check requests for the same check. The `publish_proxy_check_request() method is used to publish the proxy check requests.
@param clients [Array] of client definitions. @param check [Hash] definition.
# File lib/sensu/server/process.rb, line 907 def publish_proxy_check_requests(clients, check) client_count = clients.length splay = 0 if check[:proxy_requests][:splay] interval = check[:interval] if check[:cron] interval = determine_check_cron_time(check) end unless interval.nil? splay_coverage = check[:proxy_requests].fetch(:splay_coverage, 90) splay = interval * (splay_coverage / 100.0) / client_count end end splay_timer = 0 clients.each do |client| unless splay == 0 EM::Timer.new(splay_timer) do publish_proxy_check_request(client, check) end splay_timer += splay else publish_proxy_check_request(client, check) end end end
Relinquish a Sensu
server task. This method cancels and clears the associated task timers, those with references stored in the timers hash under `:tasks`, and removes the task from `@tasks`. The task server ID and lock are not removed from Redis, as they will be updated when another server takes reponsibility for the task, this method does not need to handle Redis connectivity issues.
@param task [String]
# File lib/sensu/server/process.rb, line 1342 def relinquish_task(task) if @tasks.include?(task) @logger.warn("relinquishing server task", :task => task) @timers[:tasks][task.to_sym].each do |timer| timer.cancel end @timers[:tasks][task.to_sym].clear @tasks.delete(task) else @logger.debug("not currently responsible for a server task", :task => task) end end
Relinquish all Sensu
server tasks, if any.
# File lib/sensu/server/process.rb, line 1356 def relinquish_tasks unless @tasks.empty? @tasks.dup.each do |task| relinquish_task(task) end else @logger.debug("not currently responsible for a server task") end end
Request a Sensu
server task election, a process to determine if the current process is to be responsible for the task. A Redis key/value is used as a central lock, using the “SETNX” Redis command to set the key/value if it does not exist, using a timestamp for the value. If the current process was able to create the key/value, it is elected, and is then responsible for the task. If the current process was not able to create the key/value, but the current timestamp value is equal to or over 30 seconds ago, the “GETSET” Redis command is used to set a new timestamp and fetch the previous value to compare them, to determine if it was set by the current process. If the current process is able to set the timestamp value, it is elected. If elected, the current process sets up the task and the associated task lock updater.
@param task [String] @yield callback/block called either after being elected and
setting up the task, or after failing to be elected.
# File lib/sensu/server/process.rb, line 1420 def request_task_election(task, &callback) @redis.setnx("lock:task:#{task}", create_lock_timestamp) do |created| if created setup_task(task, &callback) setup_task_lock_updater(task) else @redis.get("lock:task:#{task}") do |current_lock_timestamp| new_lock_timestamp = create_lock_timestamp if new_lock_timestamp - current_lock_timestamp.to_i >= 300000000 @redis.getset("lock:task:#{task}", new_lock_timestamp) do |previous_lock_timestamp| if previous_lock_timestamp == current_lock_timestamp setup_task(task, &callback) setup_task_lock_updater(task) end end else yield if block_given? end end end end end
Resume the Sensu
server process if it is currently or will soon be paused. The `retry_until_true` helper method is used to determine if the process is paused and if the Redis and transport connections are initiated and connected. If the conditions are met, `bootstrap()` will be called and true is returned to stop `retry_until_true`. If the transport has not yet been initiated, true is is returned, without calling bootstrap, as we expect bootstrap will be called after the transport initializes.
# File lib/sensu/server/process.rb, line 1612 def resume retry_until_true(1) do if @state == :paused if @redis.connected? if @transport if @transport.connected? bootstrap true end else true end end end end end
Retrieve a client (data) from Redis if it exists. If a client does not already exist, create one (a blank) using the `client_key` as the client name. Dynamically create client data can be updated using the API
(POST /clients/:client). If a client does exist and it has a client signature, the check result must have a matching signature or it is discarded. If the client does not exist, but a client signature exists, the check result must have a matching signature or it is discarded.
@param result [Hash] data. @yield [client] callback/block to be called with client data,
either retrieved from Redis, or dynamically created.
@yieldparam client [Hash]
# File lib/sensu/server/process.rb, line 665 def retrieve_client(result) client_key = result[:check][:source] || result[:client] @redis.get("client:#{client_key}") do |client_json| unless client_json.nil? client = Sensu::JSON.load(client_json) if client[:signature] if client[:signature] == result[:signature] yield(client) else @logger.warn("invalid check result signature", { :result => result, :client => client }) @logger.warn("not retrieving client from the registry", :result => result) yield(nil) end else yield(client) end else @redis.get("client:#{client_key}:signature") do |signature| if signature.nil? || signature.empty? || result[:signature] == signature client = create_client(client_key) client[:type] = "proxy" if result[:check][:source] update_client_registry(client) do yield(client) end else @logger.warn("invalid check result signature", { :result => result, :signature => signature }) yield(nil) end end end end end
Schedule a check request, using the check cron. This method determines the time until the next cron time (in seconds) and creats an EventMachine timer for the request. This method will be called after every check cron request for subsequent requests. The timer is stored in the timer hash under `:tasks`, so it can be cancelled etc. The check cron request timer object is removed from the timer hash after the request is published, to stop the timer hash from growing infinitely.
@param check [Hash] definition.
# File lib/sensu/server/process.rb, line 1003 def schedule_check_cron_request(check) cron_time = determine_check_cron_time(check) @timers[:tasks][:check_request_publisher] << EM::Timer.new(cron_time) do |timer| create_check_request_proc(check).call @timers[:tasks][:check_request_publisher].delete(timer) schedule_check_cron_request(check) end end
Schedule check requests, using the check interval. This method using an intial calculated request splay EventMachine timer and an EventMachine periodic timer for subsequent check requests. The timers are stored in the timers hash under `:tasks`, so they can be cancelled etc.
@param check [Hash] definition.
# File lib/sensu/server/process.rb, line 1030 def schedule_check_interval_requests(check) request_splay = testing? ? 0 : calculate_check_request_splay(check) interval = testing? ? 0.5 : check[:interval] @timers[:tasks][:check_request_publisher] << EM::Timer.new(request_splay) do create_check_request = create_check_request_proc(check) create_check_request.call @timers[:tasks][:check_request_publisher] << EM::PeriodicTimer.new(interval, &create_check_request) end end
Schedule check requests. This method iterates through defined checks and uses the appropriate method of check request scheduling, either with the cron syntax or a numeric interval.
@param checks [Array] of definitions.
# File lib/sensu/server/process.rb, line 1045 def schedule_checks(checks) checks.each do |check| if check[:cron] schedule_check_cron_request(check) else schedule_check_interval_requests(check) end end end
Create/return the unique Sensu
server ID for the current process.
@return [String]
# File lib/sensu/server/process.rb, line 1307 def server_id @server_id ||= random_uuid end
Set up the check request publisher. This method creates an array of check definitions, that are not standalone checks, and do not have `:publish` set to `false`. The array is provided to the `schedule_checks()` method.
# File lib/sensu/server/process.rb, line 1059 def setup_check_request_publisher @logger.debug("scheduling check requests") checks = @settings.checks.reject do |check| check[:standalone] || check[:publish] == false end schedule_checks(checks) end
Set up the check result monitor, a periodic timer to run `determine_stale_check_results()` every 30 seconds. The timer is stored in the timers hash under `:tasks`.
@param interval [Integer] to use for the check TTL result
interval.
# File lib/sensu/server/process.rb, line 1287 def setup_check_result_monitor(interval=30) @logger.debug("monitoring check results") @timers[:tasks][:check_result_monitor] << EM::PeriodicTimer.new(interval) do determine_stale_check_results(interval) end end
Set up the client monitor, a periodic timer to run `determine_stale_clients()` every 30 seconds. The timer is stored in the timers hash under `:tasks`.
# File lib/sensu/server/process.rb, line 1204 def setup_client_monitor @logger.debug("monitoring client keepalives") @timers[:tasks][:client_monitor] << EM::PeriodicTimer.new(30) do determine_stale_clients end end
Set up the Redis and Transport connection objects, `@redis` and `@transport`. This method updates the Redis on error callback to reset the in progress check result counter. This method “drys” up many instances of `setup_redis()` and `setup_transport()`, particularly in the specs.
@yield callback/block called after connecting to Redis and the
Sensu Transport.
# File lib/sensu/server/process.rb, line 62 def setup_connections setup_redis do @redis.on_error do |error| @logger.error("redis connection error", :error => error.to_s) @in_progress[:check_results] = 0 end setup_transport do yield end end end
Set up the client keepalive consumer, keeping the Sensu
client registry updated. The consumer receives JSON serialized client keepalives from the transport, parses them, and calls `update_client_registry()` with the client data to update the registry. Transport message acknowledgements are used to ensure the client registry is updated successfully. Keepalive JSON parsing errors are logged.
# File lib/sensu/server/process.rb, line 212 def setup_keepalives keepalives_pipe = "keepalives" if @settings[:sensu][:server] && @settings[:sensu][:server][:keepalives_pipe] keepalives_pipe = @settings[:sensu][:server][:keepalives_pipe] end @logger.debug("subscribing to keepalives", :pipe => keepalives_pipe) @transport.subscribe(:direct, keepalives_pipe, "keepalives", :ack => true) do |message_info, message| @logger.debug("received keepalive", :message => message) if message_size_ok?(message) begin client = Sensu::JSON.load(message) update_client_registry(client) rescue Sensu::JSON::ParseError => error @logger.error("failed to parse keepalive payload", { :message => message, :error => error.to_s }) end end EM::next_tick do @transport.ack(message_info) end end end
Set up the check result consumer. The consumer receives JSON serialized check results from the transport, parses them, and calls `process_check_result()` with the result data to be processed. Transport message acknowledgements are used to ensure that results make it to processing. The transport message acknowledgements are currently done in the next tick of the EventMachine reactor (event loop), as a flow control mechanism. Result JSON parsing errors are logged.
# File lib/sensu/server/process.rb, line 767 def setup_results results_pipe = "results" if @settings[:sensu][:server] && @settings[:sensu][:server][:results_pipe] results_pipe = @settings[:sensu][:server][:results_pipe] end @logger.debug("subscribing to results", :pipe => results_pipe) @transport.subscribe(:direct, results_pipe, "results", :ack => true) do |message_info, message| if message_size_ok?(message) begin result = Sensu::JSON.load(message) @logger.debug("received result", :result => result) process_check_result(result) rescue Sensu::JSON::ParseError => error @logger.error("failed to parse result payload", { :message => message, :error => error.to_s }) end end EM::next_tick do @transport.ack(message_info) end end end
Set up the server registry updater. A periodic timer is used to update the Sensu
server info stored in Redis. The timer is stored in the timers hash under `:run`.
# File lib/sensu/server/process.rb, line 1518 def setup_server_registry_updater update_server_registry @timers[:run] << EM::PeriodicTimer.new(10) do update_server_registry end end
Setup a Sensu
server task. Unless the current process is already responsible for the task, this method sets the tasks server ID stored in Redis to the unique random server ID for the process. If the tasks server ID is successfully updated, the task is added to `@tasks` for tracking purposes and the task setup method is called.
@param task [String] @yield callback/block called after setting up the task.
# File lib/sensu/server/process.rb, line 1320 def setup_task(task) unless @tasks.include?(task) @redis.set("task:#{task}:server", server_id) do @logger.info("i am now responsible for a server task", :task => task) @tasks << task self.send("setup_#{task}".to_sym) yield if block_given? end else @logger.debug("i am already responsible for a server task", :task => task) end end
Request Sensu
server task elections. The task list is ordered by prioity. This method works through the task list serially, increasing the election request delay as the current process becomes responsible for one or more tasks, this is to improve the initial distribution of tasks amongst Sensu
servers.
@param splay [Integer]
# File lib/sensu/server/process.rb, line 1450 def setup_task_elections(splay=10) tasks = TASKS.dup - @tasks next_task = Proc.new do task = tasks.shift if task delay = splay * @tasks.size @timers[:run] << EM::Timer.new(delay) do request_task_election(task, &next_task) end else @timers[:run] << EM::Timer.new(10) do setup_task_elections(splay) end end end next_task.call end
Set up a Sensu
server task lock updater. This method uses a periodic timer to update a task lock timestamp in Redis, every 10 seconds. If the current process fails to keep the lock timestamp updated for a task that it is responsible for, another Sensu
server will claim responsibility. This method is called after task setup.
@param task [String]
# File lib/sensu/server/process.rb, line 1396 def setup_task_lock_updater(task) @timers[:run] << EM::PeriodicTimer.new(10) do update_task_lock(task) end end
Set up Tessen
, the call home mechanism.
# File lib/sensu/server/process.rb, line 1526 def setup_tessen @tessen = Tessen.new( :settings => @settings, :logger => @logger, :redis => @redis ) @tessen.run if @tessen.enabled? end
Start the Sensu
server process, connecting to Redis, the Transport, and calling the `bootstrap()` method. Yield if a block is provided.
# File lib/sensu/server/process.rb, line 1575 def start setup_connections do bootstrap yield if block_given? end end
Stop the Sensu
server process, pausing it, completing event handling in progress, closing the Redis and transport connections, and exiting the process (exit 0). After pausing the process, the process/daemon `@state` is set to `:stopping`.
Sensu::Daemon#stop
# File lib/sensu/server/process.rb, line 1634 def stop @logger.warn("stopping") pause @state = :stopping complete_in_progress do @redis.close if @redis @transport.close if @transport super end end
Store check result data. This method stores check result data and the 21 most recent check result statuses for a client/check pair, this history is used for event context and flap detection. The check execution timestamp is also stored, to provide an indication of how recent the data is. Check output is truncated by `truncate_check_output()` before it is stored.
@param client [Hash] @param check [Hash] @yield [] callback/block called after the check data has been
stored (history, etc).
# File lib/sensu/server/process.rb, line 400 def store_check_result(client, check) @logger.debug("storing check result", :check => check) result_key = "#{client[:name]}:#{check[:name]}" history_key = "history:#{result_key}" check_truncated = truncate_check_output(check) @redis.multi @redis.sadd("result:#{client[:name]}", check[:name]) @redis.set("result:#{result_key}", Sensu::JSON.dump(check_truncated)) @redis.sadd("ttl", result_key) if check[:ttl] @redis.rpush(history_key, check[:status]) @redis.ltrim(history_key, -21, -1) if check[:status] == 0 @redis.set("#{history_key}:last_ok", check.fetch(:executed, Time.now.to_i)) end @redis.exec do yield end end
Determine the Sensu
Transport publish options for a subscription. If a subscription begins with a Transport pipe type, either “direct:” or “roundrobin:”, the subscription uses a direct Transport pipe. If a subscription does not specify a Transport pipe type, a fanout Transport pipe is used.
@param subscription [String] @param message [String] @return [Array] containing the Transport publish options:
the Transport pipe type, pipe, and the message to be published.
# File lib/sensu/server/process.rb, line 803 def transport_publish_options(subscription, message) _, raw_type = subscription.split(":", 2).reverse case raw_type when "direct", "roundrobin" [:direct, subscription, message] else [:fanout, subscription, message] end end
Truncate check output. Metric checks (`“type”: “metric”`), or checks with `“truncate_output”: true`, have their output truncated to a single line and a maximum character length of 255 by default. The maximum character length can be change by the `“truncate_output_length”` check definition attribute.
@param check [Hash] @return [Hash] check with truncated output.
# File lib/sensu/server/process.rb, line 365 def truncate_check_output(check) if check[:truncate_output] || (check[:type] == METRIC_CHECK_TYPE && check[:truncate_output] != false) begin output_lines = check[:output].split("\n") rescue ArgumentError utf8_output = check[:output].encode("UTF-8", "binary", { :invalid => :replace, :undef => :replace, :replace => "" }) output_lines = utf8_output.split("\n") end output = output_lines.first || check[:output] truncate_output_length = check.fetch(:truncate_output_length, 255) if output_lines.length > 1 || output.length > truncate_output_length output = output[0..truncate_output_length] + "\n..." end check.merge(:output => output) else check end end
Unsubscribe from transport subscriptions (all of them). This method is called when there are issues with connectivity, or the process is stopping.
# File lib/sensu/server/process.rb, line 1538 def unsubscribe @logger.warn("unsubscribing from keepalive and result queues") @transport.unsubscribe if @transport end
Update the Sensu
client registry, stored in Redis. Sensu
client data is used to provide additional event context and enable agent health monitoring.
To enable silencing individual clients, per-client subscriptions (`client:$CLIENT_NAME`) are added to client subscriptions automatically.
The client registry supports client signatures, unique string identifiers used for keepalive and result source verification. If a client has a signature, all further registry updates for the client must have the same signature. A client can begin to use a signature if one was not previously configured. JSON serialization is used for the stored client data.
@param client [Hash] @yield [success] passes success status to optional
callback/block.
@yieldparam success [TrueClass,FalseClass] indicating if the
client registry update was a success or the client data was discarded due to client signature mismatch.
# File lib/sensu/server/process.rb, line 148 def update_client_registry(client) @logger.debug("updating client registry", :client => client) client_key = "client:#{client[:name]}" client[:subscriptions] = (client[:subscriptions] + [client_key]).uniq signature_key = "#{client_key}:signature" @redis.setnx(signature_key, client[:signature]) do |created| process_client_registration(client) if created @redis.get(signature_key) do |signature| if (signature.nil? || signature.empty?) && client[:signature] @redis.set(signature_key, client[:signature]) end if signature.nil? || signature.empty? || client[:signature] == signature @redis.multi @redis.set(client_key, Sensu::JSON.dump(client)) @redis.sadd("clients", client[:name]) @redis.exec do yield(true) if block_given? end else @logger.warn("invalid client signature", { :client => client, :signature => signature }) @logger.warn("not updating client in the registry", :client => client) yield(false) if block_given? end end end end
Update the event registry, stored in Redis. This method determines if event data warrants in the creation or update of event data in the registry. If a check `:status` is not `0`, or it has been flapping, an event is created/updated in the registry. If the event `:action` is `:resolve`, the event is removed (resolved) from the registry. If the previous conditions are not met and check `:type` is `metric`, the registry is not updated, but further event processing is required (`yield(true)`). JSON serialization is used when storing data in the registry.
@param event [Hash] @yield callback [event] callback/block called after the event
registry has been updated.
@yieldparam process [TrueClass, FalseClass] indicating if the
event requires further processing.
# File lib/sensu/server/process.rb, line 558 def update_event_registry(event) client_name = event[:client][:name] if event[:check][:status] != 0 || (event[:action] == :flapping && event[:check][:force_resolve] != true) @redis.hset("events:#{client_name}", event[:check][:name], Sensu::JSON.dump(event)) do yield(true) end elsif event[:action] == :resolve && (event[:check][:auto_resolve] != false || event[:check][:force_resolve]) || (event[:action] == :flapping && event[:check][:force_resolve]) @redis.hdel("events:#{client_name}", event[:check][:name]) do yield(true) end elsif event[:check][:type] == METRIC_CHECK_TYPE yield(true) else yield(false) end end
Update the Sensu
server registry, stored in Redis. This method adds the local/current Sensu
server info to the registry, including its id, hostname, address, its server tasks, and some metrics. Sensu
server registry entries expire in 30 seconds unless updated.
@yield [success] passes success status to optional
callback/block.
@yieldparam success [TrueClass,FalseClass] indicating if the
server registry update was a success.
# File lib/sensu/server/process.rb, line 1478 def update_server_registry @logger.debug("updating the server registry") process_cpu_times do |cpu_user, cpu_system, _, _| sensu = RELEASE_INFO.merge( :settings => { :hexdigest => @settings.hexdigest } ) tessen = @settings[:tessen] || {} tessen_enabled = tessen.fetch(:enabled, false) info = { :id => server_id, :hostname => system_hostname, :address => system_address, :tasks => @tasks, :metrics => { :cpu => { :user => cpu_user, :system => cpu_system } }, :sensu => sensu, :tessen => { :enabled => tessen_enabled }, :timestamp => Time.now.to_i } @redis.sadd("servers", server_id) server_key = "server:#{server_id}" @redis.set(server_key, Sensu::JSON.dump(info)) do @redis.expire(server_key, 30) @logger.info("updated server registry", :server => info) yield(true) if block_given? end end end
Updates a Sensu
server task lock timestamp. The current task server ID is retrieved from Redis and compared with the server ID of the current process to determine if it is still responsible for the task. If the current process is still responsible, the task lock timestamp is updated. If the current process is no longer responsible, `relinquish_task()` is called for cleanup.
@param task [String]
# File lib/sensu/server/process.rb, line 1375 def update_task_lock(task) @redis.get("task:#{task}:server") do |current_server_id| if current_server_id == server_id @redis.set("lock:task:#{task}", create_lock_timestamp) do @logger.debug("updated task lock timestamp", :task => task) end else @logger.warn("another sensu server is responsible for the task", :task => task) relinquish_task(task) end end end