Class: Sensu::Server::Process

Inherits:
Object
  • Object
show all
Includes:
Daemon, Filter, Handle, Mutate
Defined in:
lib/sensu/server/process.rb

Constant Summary collapse

METRIC_CHECK_TYPE = 
"metric".freeze
EVENT_FLAPPING_ACTION = 
"flapping".freeze
DEFAULT_HANDLER_NAME = 
"default".freeze

Constants included from Filter

Filter::EVAL_PREFIX

Instance Attribute Summary collapse

Attributes included from Daemon

#start_time

Class Method Summary collapse

Instance Method Summary collapse

Methods included from Handle

#handle_event, #handler_error, #handler_extension, #handler_type_router, #pipe_handler, #tcp_handler, #transport_handler, #udp_handler

Methods included from Mutate

#mutate_event, #mutator_callback, #mutator_extension, #pipe_mutator

Methods included from Filter

#action_subdued?, #check_request_subdued?, #eval_attribute_value, #event_filter, #event_filtered?, #filter_attributes_match?, #filter_event, #handle_action?, #handle_severity?, #handler_subdued?, #handling_disabled?, #process_eval_string, #subdue_days?, #subdue_exception?, #subdue_time?

Methods included from Daemon

#load_extensions, #load_settings, #log_concerns, #print_settings, #setup_logger, #setup_process, #setup_redis, #setup_signal_traps, #setup_transport

Methods included from Utilities

#deep_merge, #find_attribute_value, #random_uuid, #redact_sensitive, #retry_until_true, #substitute_tokens, #testing?

Constructor Details

#initialize(options = {}) ⇒ Process

Override Daemon initialize() to support Sensu server leader election and the handling event count.

Parameters:

  • options (Hash) (defaults to: {}) 


39
40
41
42
43
44
 
# File 'lib/sensu/server/process.rb', line 39

def initialize(options={})
  super
  @is_leader = false
  @timers[:leader] = Array.new
  @handling_event_count = 0
end

Instance Attribute Details

#handling_event_count ⇒ Object (readonly)

Returns the value of attribute handling_event_count.



14
15
16
 
# File 'lib/sensu/server/process.rb', line 14

def handling_event_count
  @handling_event_count
end

#is_leader ⇒ Object (readonly)

Returns the value of attribute is_leader.



14
15
16
 
# File 'lib/sensu/server/process.rb', line 14

def is_leader
  @is_leader
end

Class Method Details

.run(options = {}) ⇒ Object

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).

Parameters:

  • options (Hash) (defaults to: {}) 


27
28
29
30
31
32
33
 
# File 'lib/sensu/server/process.rb', line 27

def self.run(options={})
  server = self.new(options)
  EM::run do
    server.start
    server.setup_signal_traps
  end
end

Instance Method Details

#aggregate_check_result(client, check) ⇒ Object

Add a check result to an aggregate. A check aggregate uses the check ‘:name` and the `:issued` timestamp as its unique identifier. An aggregate uses several counters: the total number of results in the aggregate, and a counter for each check severity (ok, warning, etc). Check output is also stored, to be summarized to aid in identifying outliers for a check execution across a number of Sensu clients. JSON serialization is used for storing check result data.

Parameters:

  • client (Hash)
  • check (Hash) 


285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
 
# File 'lib/sensu/server/process.rb', line 285

def aggregate_check_result(client, check)
  @logger.debug("adding check result to aggregate", {
    :client => client,
    :check => check
  })
  result_set = "#{check[:name]}:#{check[:issued]}"
  result_data = Sensu::JSON.dump(:output => check[:output], :status => check[:status])
  @redis.multi
  @redis.hset("aggregation:#{result_set}", client[:name], result_data)
  SEVERITIES.each do |severity|
    @redis.hsetnx("aggregate:#{result_set}", severity, 0)
  end
  severity = (SEVERITIES[check[:status]] || "unknown")
  @redis.hincrby("aggregate:#{result_set}", severity, 1)
  @redis.hincrby("aggregate:#{result_set}", "total", 1)
  @redis.sadd("aggregates:#{check[:name]}", check[:issued])
  @redis.sadd("aggregates", check[:name])
  @redis.exec
end

#become_the_leader ⇒ Object

Become the Sensu server leader, responsible for specific duties (‘leader_duties()`). Unless the current process is already the leader, this method sets the leader ID stored in Redis to the unique random leader ID for the process. If the leader ID in Redis is successfully updated, `@is_leader` is set to true and `leader_duties()` is called to begin the tasks/duties of the Sensu server leader.



931
932
933
934
935
936
937
938
939
940
941
 
# File 'lib/sensu/server/process.rb', line 931

def become_the_leader
  unless @is_leader
    @redis.set("leader", leader_id) do
      @logger.info("i am now the leader")
      @is_leader = true
      leader_duties
    end
  else
    @logger.debug("i am already the leader")
  end
end

#bootstrap ⇒ Object

Bootstrap the Sensu server process, setting up the keepalive and check result consumers, and attemping to become the leader to carry out its duties. This method sets the process/daemon ‘@state` to `:running`.



1066
1067
1068
1069
1070
1071
 
# File 'lib/sensu/server/process.rb', line 1066

def bootstrap
  setup_keepalives
  setup_results
  setup_leader_monitor
  @state = :running
end

#calculate_check_execution_splay(check) ⇒ Object

Calculate a check execution splay, taking into account the current time and the execution interval to ensure it’s consistent between process restarts.

Parameters:

  • check (Hash) —

    definition.



655
656
657
658
659
 
# File 'lib/sensu/server/process.rb', line 655

def calculate_check_execution_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

#check_flapping?(stored_event, check) ⇒ TrueClass, FalseClass

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

Parameters:

  • stored_event (Hash)
  • check (Hash)

Returns:

  • (TrueClass, FalseClass) 


407
408
409
410
411
412
413
414
415
416
417
418
 
# File 'lib/sensu/server/process.rb', line 407

def check_flapping?(stored_event, check)
  if check.has_key?(:low_flap_threshold) && check.has_key?(:high_flap_threshold)
    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
    false
  end
end

#check_history(client, check) {|history, total_state_change| ... } ⇒ Object

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

Parameters:

  • client (Hash)
  • check (Hash)

Yields:

  • (history, total_state_change) —

    callback/block to call with the check history and calculated total state change value.

Yield Parameters:

  • history (Array) —

    containing the last 21 check result exit status codes.

  • total_state_change (Float) —

    percentage for the check history (exit status codes).



368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
 
# File 'lib/sensu/server/process.rb', line 368

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
    yield(history, total_state_change)
  end
end

#complete_event_handling { ... } ⇒ Object

Complete event handling currently in progress. The ‘:handling_event_count` is used to determine if event handling is complete, when it is equal to `0`. The provided callback is called when handling is complete.

Yields:

  • callback/block to call when event handling is

    complete.



1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
 
# File 'lib/sensu/server/process.rb', line 1050

def complete_event_handling
  @logger.info("completing event handling in progress", {
    :handling_event_count => @handling_event_count
  })
  retry_until_true do
    if @handling_event_count == 0
      yield
      true
    end
  end
end

#create_client(name) {|client| ... } ⇒ Object

Create a blank client (data) and add it to the client registry. 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, ‘:keepalives` is set to `false`.

Parameters:

  • name (Hash) —

    to use for the client.

Yields:

  • (client) —

    callback/block to be called with the dynamically created client data.

Yield Parameters:

  • client (Hash) 


487
488
489
490
491
492
493
494
495
496
497
498
 
# File 'lib/sensu/server/process.rb', line 487

def create_client(name)
  client = {
    :name => name,
    :address => "unknown",
    :subscriptions => [],
    :keepalives => false,
    :version => VERSION
  }
  update_client_registry(client) do
    yield(client)
  end
end

#create_client_registration_event(client) ⇒ Object

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.

Parameters:

  • client (Hash) —

    definition.



92
93
94
95
96
97
98
99
100
101
102
 
# File 'lib/sensu/server/process.rb', line 92

def create_client_registration_event(client)
  event = {
    :id => random_uuid,
    :client => client,
    :check => create_registration_check(client),
    :occurrences => 1,
    :action => :create,
    :timestamp => Time.now.to_i
  }
  process_event(event)
end

#create_keepalive_check(client) ⇒ Object

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.



741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
 
# File 'lib/sensu/server/process.rb', line 741

def create_keepalive_check(client)
  check = {
    :thresholds => {
      :warning => 120,
      :critical => 180
    }
  }
  if @settings.handler_exists?(:keepalive)
    check[:handler] = "keepalive"
  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)
end

#create_lock_timestamp ⇒ Integer

Create a lock timestamp (integer), current time including milliseconds. This method is used by Sensu server leader election.

Returns:

  • (Integer) 


912
913
914
 
# File 'lib/sensu/server/process.rb', line 912

def create_lock_timestamp
  (Time.now.to_f * 1000).to_i
end

#create_registration_check(client) ⇒ Object

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`, `:status`, `:issued`, and `:executed` values are always overridden to guard against an invalid definition.



69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
 
# File 'lib/sensu/server/process.rb', line 69

def create_registration_check(client)
  check = {:handler => "registration"}
  if client.has_key?(:registration)
    check = deep_merge(check, client[:registration])
  end
  timestamp = Time.now.to_i
  overrides = {
    :name => "registration",
    :output => "new client registration",
    :status => 1,
    :issued => timestamp,
    :executed => timestamp
  }
  check.merge(overrides)
end

#derive_handlers(handler_list, depth = 0) ⇒ Array

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.

Parameters:

  • handler_list (Array)
  • depth (Integer) (defaults to: 0) —

    of handler set expansion.

Returns:

  • (Array) 


219
220
221
222
223
224
225
226
227
228
229
230
231
232
 
# File 'lib/sensu/server/process.rb', line 219

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_stale_check_results ⇒ Object

Determine stale check results, those that have not executed in a specified amount of time (check TTL). This method iterates through the client registry and check results for checks with a defined TTL value (in seconds). If a check result has a defined TTL, the time since last check execution (in seconds) is calculated. 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.



818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
 
# File 'lib/sensu/server/process.rb', line 818

def determine_stale_check_results
  @logger.info("determining stale check results")
  @redis.smembers("clients") do |clients|
    clients.each do |client_name|
      @redis.smembers("result:#{client_name}") do |checks|
        checks.each do |check_name|
          result_key = "#{client_name}:#{check_name}"
          @redis.get("result:#{result_key}") do |result_json|
            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]
                check[:output] = "Last check execution was "
                check[:output] << "#{time_since_last_execution} seconds ago"
                check[:status] = 1
                publish_check_result(client_name, check)
              end
            end
          end
        end
      end
    end
  end
end

#determine_stale_clients ⇒ Object

Determine stale clients, those that have not sent a keepalive in a specified amount of time (thresholds). This method iterates through the client registry, creating a keepalive check definition with the ‘create_keepalive_check()` method, containing client specific staleness 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. A check result is published for every client in the registry.



769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
 
# File 'lib/sensu/server/process.rb', line 769

def determine_stale_clients
  @logger.info("determining stale clients")
  @redis.smembers("clients") do |clients|
    clients.each do |client_name|
      @redis.get("client:#{client_name}") do |client_json|
        unless client_json.nil?
          client = Sensu::JSON.load(client_json)
          next if client[:keepalives] == false
          check = 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
          publish_check_result(client[:name], check)
        end
      end
    end
  end
end

#event_bridges(event) ⇒ Object

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.

Parameters:

  • event (Hash) 


263
264
265
266
267
268
269
270
271
272
 
# File 'lib/sensu/server/process.rb', line 263

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

#expand_handler_sets(handler, depth = 0) ⇒ Array, ...

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.

Parameters:

  • handler (Hash) —

    definition.

  • depth (Integer) (defaults to: 0) —

    of the expansion.

Returns:

  • (Array, Hash, Nil) 


197
198
199
200
201
202
203
204
205
206
207
208
 
# File 'lib/sensu/server/process.rb', line 197

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

#leader_duties ⇒ Object

Set up the leader duties, tasks only performed by a single Sensu server at a time. The duties include publishing check requests, monitoring for stale clients, and pruning check result aggregations.



900
901
902
903
904
905
 
# File 'lib/sensu/server/process.rb', line 900

def leader_duties
  setup_check_request_publisher
  setup_client_monitor
  setup_check_result_monitor
  setup_check_result_aggregation_pruner
end

#leader_id ⇒ String

Create/return the unique Sensu server leader ID for the current process.

Returns:

  • (String) 


920
921
922
 
# File 'lib/sensu/server/process.rb', line 920

def leader_id
  @leader_id ||= random_uuid
end

#pause ⇒ Object

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, and the references are cleared. The Sensu server will unsubscribe from all transport subscriptions, resign as leader (if currently the leader), then set the process/daemon ‘@state` to `:paused`.



1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
 
# File 'lib/sensu/server/process.rb', line 1088

def pause
  unless @state == :pausing || @state == :paused
    @state = :pausing
    @timers[:run].each do |timer|
      timer.cancel
    end
    @timers[:run].clear
    unsubscribe
    resign_as_leader
    @state = :paused
  end
end

#process_check_result(result) ⇒ Object

Process a check result, storing its data, inspecting its contents, and taking the appropriate actions (eg. update 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.

Parameters:

  • result (Hash) —

    data.



546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
 
# File 'lib/sensu/server/process.rb', line 546

def process_check_result(result)
  @logger.debug("processing result", :result => result)
  retrieve_client(result) do |client|
    check = case
    when @settings.check_exists?(result[:check][:name]) && !result[:check][:standalone]
      @settings[:checks][result[:check][:name]].merge(result[:check])
    else
      result[:check]
    end
    aggregate_check_result(client, check) if check[:aggregate]
    store_check_result(client, check) do
      check_history(client, check) do |history, total_state_change|
        check[:history] = history
        check[:total_state_change] = total_state_change
        update_event_registry(client, check) do |event|
          process_event(event)
        end
      end
    end
  end
end

#process_client_registration(client) ⇒ Object

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()`).

Parameters:

  • client (Hash) —

    definition.



111
112
113
114
115
 
# File 'lib/sensu/server/process.rb', line 111

def process_client_registration(client)
  if @settings.handler_exists?("registration") || client[:registration]
    create_client_registration_event(client)
  end
end

#process_event(event) ⇒ Object

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 ‘@handling_event_count` is incremented by `1`, for each event handler chain (filter -> mutate -> handle).

Parameters:

  • event (Hash) 


242
243
244
245
246
247
248
249
250
251
252
253
254
255
 
# File 'lib/sensu/server/process.rb', line 242

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|
    @handling_event_count += 1
    filter_event(handler, event) do |event|
      mutate_event(handler, event) do |event_data|
        handle_event(handler, event_data)
      end
    end
  end
end

#prune_check_result_aggregations ⇒ Object

Prune check result aggregations (aggregates). Sensu only stores the 20 latest aggregations for a check, to keep the amount of data stored to a minimum.



857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
 
# File 'lib/sensu/server/process.rb', line 857

def prune_check_result_aggregations
  @logger.info("pruning check result aggregations")
  @redis.smembers("aggregates") do |checks|
    checks.each do |check_name|
      @redis.smembers("aggregates:#{check_name}") do |aggregates|
        if aggregates.length > 20
          aggregates.sort!
          aggregates.take(aggregates.length - 20).each do |check_issued|
            result_set = "#{check_name}:#{check_issued}"
            @redis.multi
            @redis.srem("aggregates:#{check_name}", check_issued)
            @redis.del("aggregate:#{result_set}")
            @redis.del("aggregation:#{result_set}")
            @redis.exec do
              @logger.debug("pruned aggregation", {
                :check => {
                  :name => check_name,
                  :issued => check_issued
                }
              })
            end
          end
        end
      end
    end
  end
end

#publish_check_request(check) ⇒ Object

Publish a check request to the transport. A check request is composed of a check ‘:name`, an `:issued` timestamp, a check `:command` if available, and a check `:extension if available. 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.

Parameters:

  • check (Hash) —

    definition.



624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
 
# File 'lib/sensu/server/process.rb', line 624

def publish_check_request(check)
  payload = {
    :name => check[:name],
    :issued => Time.now.to_i
  }
  payload[:command] = check[:command] if check.has_key?(:command)
  payload[:source] = check[:source] if check.has_key?(:source)
  payload[:extension] = check[:extension] if check.has_key?(:extension)
  @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_check_result(client_name, check) ⇒ Object

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.

Parameters:

  • client_name (String)
  • check (Hash) 


713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
 
# File 'lib/sensu/server/process.rb', line 713

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

#request_leader_election ⇒ Object

Request a leader election, a process to determine if the current process is the Sensu server leader, with its own/unique duties. 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 the leader, and must do the duties of the leader. 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 becomes the leader.



997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
 
# File 'lib/sensu/server/process.rb', line 997

def request_leader_election
  @redis.setnx("lock:leader", create_lock_timestamp) do |created|
    if created
      become_the_leader
    else
      @redis.get("lock:leader") do |current_lock_timestamp|
        new_lock_timestamp = create_lock_timestamp
        if new_lock_timestamp - current_lock_timestamp.to_i >= 30000
          @redis.getset("lock:leader", new_lock_timestamp) do |previous_lock_timestamp|
            if previous_lock_timestamp == current_lock_timestamp
              become_the_leader
            end
          end
        end
      end
    end
  end
end

#resign_as_leader ⇒ Object

Resign as leader, if the current process is the Sensu server leader. This method cancels and clears the leader timers, those with references stored in the timers hash under ‘:leader`, and `@is_leader` is set to `false`. The leader ID and leader lock are not removed from Redis, as they will be updated when another server is elected to be the leader, this method does not need to handle Redis connectivity issues.



950
951
952
953
954
955
956
957
958
959
960
961
 
# File 'lib/sensu/server/process.rb', line 950

def resign_as_leader
  if @is_leader
    @logger.warn("resigning as leader")
    @timers[:leader].each do |timer|
      timer.cancel
    end
    @timers[:leader].clear
    @is_leader = false
  else
    @logger.debug("not currently the leader")
  end
end

#resume ⇒ Object

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 connected. If the conditions are met, `bootstrap()` will be called and true is returned to stop `retry_until_true`.



1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
 
# File 'lib/sensu/server/process.rb', line 1107

def resume
  retry_until_true(1) do
    if @state == :paused
      if @redis.connected? && @transport.connected?
        bootstrap
        true
      end
    end
  end
end

#retrieve_client(result) {|client| ... } ⇒ Object

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.

Parameters:

  • result (Hash) —

    data.

Yields:

  • (client) —

    callback/block to be called with client data, either retrieved from Redis, or dynamically created.

Yield Parameters:

  • client (Hash) 


511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
 
# File 'lib/sensu/server/process.rb', line 511

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)
        end
      else
        yield(client)
      end
    else
      create_client(client_key) do |client|
        yield(client)
      end
    end
  end
end

#schedule_check_executions(checks) ⇒ Object

Schedule check executions, using EventMachine periodic timers, using a calculated execution splay. The timers are stored in the timers hash under ‘:leader`, as check request publishing is a task for only the Sensu server leader, so they can be cancelled etc. Check requests are not published if subdued.

Parameters:

  • checks (Array) —

    of definitions.



668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
 
# File 'lib/sensu/server/process.rb', line 668

def schedule_check_executions(checks)
  checks.each do |check|
    create_check_request = Proc.new do
      unless check_request_subdued?(check)
        publish_check_request(check)
      else
        @logger.info("check request was subdued", :check => check)
      end
    end
    execution_splay = testing? ? 0 : calculate_check_execution_splay(check)
    interval = testing? ? 0.5 : check[:interval]
    @timers[:leader] << EM::Timer.new(execution_splay) do
      create_check_request.call
      @timers[:leader] << EM::PeriodicTimer.new(interval, &create_check_request)
    end
  end
end

#setup_check_request_publisher ⇒ Object

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 of check definitions includes those from standard checks and extensions (with a defined execution `:interval`). The array is provided to the `schedule_check_executions()` method.



692
693
694
695
696
697
698
699
700
701
 
# File 'lib/sensu/server/process.rb', line 692

def setup_check_request_publisher
  @logger.debug("scheduling check requests")
  standard_checks = @settings.checks.reject do |check|
    check[:standalone] || check[:publish] == false
  end
  extension_checks = @extensions.checks.reject do |check|
    check[:standalone] || check[:publish] == false || !check[:interval].is_a?(Integer)
  end
  schedule_check_executions(standard_checks + extension_checks)
end

#setup_check_result_aggregation_pruner ⇒ Object

Set up the check result aggregation pruner, using periodic timer to run ‘prune_check_result_aggregations()` every 20 seconds. The timer is stored in the timers hash under `:leader`.



889
890
891
892
893
894
 
# File 'lib/sensu/server/process.rb', line 889

def setup_check_result_aggregation_pruner
  @logger.debug("pruning check result aggregations")
  @timers[:leader] << EM::PeriodicTimer.new(20) do
    prune_check_result_aggregations
  end
end

#setup_check_result_monitor ⇒ Object

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 `:leader`.



847
848
849
850
851
852
 
# File 'lib/sensu/server/process.rb', line 847

def setup_check_result_monitor
  @logger.debug("monitoring check results")
  @timers[:leader] << EM::PeriodicTimer.new(30) do
    determine_stale_check_results
  end
end

#setup_client_monitor ⇒ Object

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 `:leader`.



803
804
805
806
807
808
 
# File 'lib/sensu/server/process.rb', line 803

def setup_client_monitor
  @logger.debug("monitoring client keepalives")
  @timers[:leader] << EM::PeriodicTimer.new(30) do
    determine_stale_clients
  end
end

#setup_connections { ... } ⇒ Object

Set up the Redis and Transport connection objects, ‘@redis` and `@transport`. This method “drys” up many instances of `setup_redis()` and `setup_transport()`.

Yields:

  • callback/block called after connecting to Redis and the Sensu Transport.



52
53
54
55
56
57
58
 
# File 'lib/sensu/server/process.rb', line 52

def setup_connections
  setup_redis do
    setup_transport do
      yield
    end
  end
end

#setup_keepalives ⇒ Object

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.



169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
 
# File 'lib/sensu/server/process.rb', line 169

def setup_keepalives
  @logger.debug("subscribing to keepalives")
  @transport.subscribe(:direct, "keepalives", "keepalives", :ack => true) do |message_info, message|
    @logger.debug("received keepalive", :message => message)
    begin
      client = Sensu::JSON.load(message)
      update_client_registry(client) do
        @transport.ack(message_info)
      end
    rescue Sensu::JSON::ParseError => error
      @logger.error("failed to parse keepalive payload", {
        :message => message,
        :error => error.to_s
      })
      @transport.ack(message_info)
    end
  end
end

#setup_leader_monitor ⇒ Object

Set up the leader monitor. A one-time timer is used to run ‘request_leader_exection()` in 2 seconds. A periodic timer is used to update the leader lock timestamp if the current process is the leader, or to run `request_leader_election(), every 10 seconds. The timers are stored in the timers hash under `:run`.



1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
 
# File 'lib/sensu/server/process.rb', line 1022

def setup_leader_monitor
  @timers[:run] << EM::Timer.new(2) do
    request_leader_election
  end
  @timers[:run] << EM::PeriodicTimer.new(10) do
    if @is_leader
      update_leader_lock
    else
      request_leader_election
    end
  end
end

#setup_results ⇒ Object

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.



576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
 
# File 'lib/sensu/server/process.rb', line 576

def setup_results
  @logger.debug("subscribing to results")
  @transport.subscribe(:direct, "results", "results", :ack => true) do |message_info, 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
    EM::next_tick do
      @transport.ack(message_info)
    end
  end
end

#start ⇒ Object

Start the Sensu server process, connecting to Redis, the transport, and calling the ‘bootstrap()` method.



1075
1076
1077
1078
1079
 
# File 'lib/sensu/server/process.rb', line 1075

def start
  setup_connections do
    bootstrap
  end
end

#stop ⇒ Object

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`.



1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
 
# File 'lib/sensu/server/process.rb', line 1123

def stop
  @logger.warn("stopping")
  pause
  @state = :stopping
  complete_event_handling do
    @redis.close if @redis
    @transport.close if @transport
    super
  end
end

#store_check_result(client, check) { ... } ⇒ Object

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.

stored (history, etc).

Parameters:

  • client (Hash)
  • check (Hash)

Yields:

  • callback/block called after the check data has been



337
338
339
340
341
342
343
344
345
346
347
348
349
350
 
# File 'lib/sensu/server/process.rb', line 337

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.rpush(history_key, check[:status])
  @redis.ltrim(history_key, -21, -1)
  @redis.exec do
    yield
  end
end

#transport_publish_options(subscription, message) ⇒ Array

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.

Parameters:

  • subscription (String)

Returns:

  • (Array) —

    containing the transport publish options: the transport pipe type, pipe, and the message to be published.



605
606
607
608
609
610
611
612
613
 
# File 'lib/sensu/server/process.rb', line 605

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(check) ⇒ Hash

Truncate check output. For metric checks, (‘“type”: “metric”`), check output is truncated to a single line and a maximum of 255 characters. Check output is currently left unmodified for standard checks.

Parameters:

  • check (Hash)

Returns:

  • (Hash) —

    check with truncated output.



312
313
314
315
316
317
318
319
320
321
322
323
324
 
# File 'lib/sensu/server/process.rb', line 312

def truncate_check_output(check)
  case check[:type]
  when METRIC_CHECK_TYPE
    output_lines = check[:output].split("\n")
    output = output_lines.first || check[:output]
    if output_lines.length > 1 || output.length > 255
      output = output[0..255] + "\n..."
    end
    check.merge(:output => output)
  else
    check
  end
end

#unsubscribe ⇒ Object

Unsubscribe from transport subscriptions (all of them). This method is called when there are issues with connectivity, or the process is stopping.



1038
1039
1040
1041
 
# File 'lib/sensu/server/process.rb', line 1038

def unsubscribe
  @logger.warn("unsubscribing from keepalive and result queues")
  @transport.unsubscribe if @transport
end

#update_client_registry(client) {|success| ... } ⇒ Object

Update the Sensu client registry, stored in Redis. Sensu client data is used to provide additional event context and enable agent health monitoring. 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.

Parameters:

  • client (Hash)

Yields:

  • (success) —

    passes success status to optional callback/block.

Yield Parameters:

  • success (TrueClass, FalseClass) —

    indicating if the client registry update was a success or the client data was discarded due to client signature mismatch.



133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
 
# File 'lib/sensu/server/process.rb', line 133

def update_client_registry(client)
  @logger.debug("updating client registry", :client => client)
  client_key = "client:#{client[:name]}"
  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_event_registry(client, check) {|event| ... } ⇒ Object

Update the event registry, stored in Redis. This method determines if check data results in the creation or update of event data in the registry. Existing event data for a client/check pair is fetched, used in conditionals and the composition of the new event data. If a check ‘:status` is not `0`, or it has been flapping, an event is created/updated in the registry. If there was existing event data, but the check `:status` is now `0`, the event is removed (resolved) from the registry. If the previous conditions are not met, and check `:type` is `metric` and the `:status` is `0`, the event registry is not updated, but the provided callback is called with the event data. All event data is sent to event bridge extensions, including events that do not normally produce an action. JSON serialization is used when storing data in the registry.

Parameters:

  • client (Hash)
  • check (Hash)

Yield Parameters:

  • event (Hash) 


442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
 
# File 'lib/sensu/server/process.rb', line 442

def update_event_registry(client, check)
  @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,
      :action => (flapping ? :flapping : :create),
      :timestamp => Time.now.to_i
    }
    if check[:status] != 0 || flapping
      if stored_event && check[:status] == stored_event[:check][:status]
        event[:occurrences] = stored_event[:occurrences] + 1
      end
      @redis.hset("events:#{client[:name]}", check[:name], Sensu::JSON.dump(event)) do
        yield(event)
      end
    elsif stored_event
      event[:occurrences] = stored_event[:occurrences]
      event[:action] = :resolve
      unless check[:auto_resolve] == false && !check[:force_resolve]
        @redis.hdel("events:#{client[:name]}", check[:name]) do
          yield(event)
        end
      end
    elsif check[:type] == METRIC_CHECK_TYPE
      yield(event)
    end
    event_bridges(event)
  end
end

#update_leader_lock ⇒ Object

Updates the Sensu server leader lock timestamp. The current leader ID is retrieved from Redis and compared with the leader ID of the current process to determine if it is still the Sensu server leader. If the current process is still the leader, the leader lock timestamp is updated. If the current process is no longer the leader (regicide), ‘resign_as_leader()` is called for cleanup, so there is not more than one leader.



971
972
973
974
975
976
977
978
979
980
981
982
 
# File 'lib/sensu/server/process.rb', line 971

def update_leader_lock
  @redis.get("leader") do |current_leader_id|
    if current_leader_id == leader_id
      @redis.set("lock:leader", create_lock_timestamp) do
        @logger.debug("updated leader lock timestamp")
      end
    else
      @logger.warn("another sensu server has been elected as leader")
      resign_as_leader
    end
  end
end