executive.rb in flapjack-0.6.38

- old
+ new

@@ -43,11 +43,14 @@
       @filters << Flapjack::Filters::UnscheduledMaintenance.new(options)
       @filters << Flapjack::Filters::DetectMassClientFailures.new(options)
       @filters << Flapjack::Filters::Delays.new(options)
       @filters << Flapjack::Filters::Acknowledgement.new(options)
 
-      @boot_time = Time.now
+      @boot_time    = Time.now
+      @fqdn         = `/bin/hostname -f`.chomp
+      @pid          = Process.pid
+      @instance_id  = "#{@fqdn}:#{@pid}"
 
       # FIXME: all of the below keys assume there is only ever one executive running;
       # we could generate a fuid and save it to disk, and prepend it from that
       # point on...
 
@@ -59,10 +62,16 @@
         @redis.hset('event_counters', 'all', 0)
         @redis.hset('event_counters', 'ok', 0)
         @redis.hset('event_counters', 'failure', 0)
         @redis.hset('event_counters', 'action', 0)
       end
+
+      @redis.zadd('executive_instances', @boot_time.to_i, @instance_id)
+      @redis.hset("event_counters:#{@instance_id}", 'all', 0)
+      @redis.hset("event_counters:#{@instance_id}", 'ok', 0)
+      @redis.hset("event_counters:#{@instance_id}", 'failure', 0)
+      @redis.hset("event_counters:#{@instance_id}", 'action', 0)
     end
 
     def main
       setup
 
@@ -71,10 +80,13 @@
       until should_quit?
         @logger.info("Waiting for event...")
         event = Flapjack::Data::Event.next(:persistence => @redis)
         process_event(event) unless event.nil?
       end
+
+      @redis.empty! if @redis
+
       @logger.info("Exiting main loop.")
     end
 
     # this must use a separate connection to the main Executive one, as it's running
     # from a different fiber while the main one is blocking.
@@ -121,10 +133,11 @@
 
     def update_keys(event, entity_check)
       result    = { :skip_filters => false }
       timestamp = Time.now.to_i
       @event_count = @redis.hincrby('event_counters', 'all', 1)
+      @event_count = @redis.hincrby("event_counters:#{@instance_id}", 'all', 1)
 
       # FIXME skip if entity_check.nil?
 
       # FIXME: validate that the event is sane before we ever get here
       # FIXME: create an event if there is dodgy data
@@ -135,12 +148,14 @@
         # Track when we last saw an event for a particular entity:check pair
         entity_check.last_update = timestamp
 
         if event.ok?
           @redis.hincrby('event_counters', 'ok', 1)
+          @redis.hincrby("event_counters:#{@instance_id}", 'ok', 1)
         elsif event.failure?
           @redis.hincrby('event_counters', 'failure', 1)
+          @redis.hincrby("event_counters:#{@instance_id}", 'failure', 1)
           @redis.hset('unacknowledged_failures', @event_count, event.id)
         end
 
         event.previous_state = entity_check.state
         @logger.info("No previous state for event #{event.id}") if event.previous_state.nil?
@@ -165,10 +180,11 @@
 
       # Action events represent human or automated interaction with Flapjack
       when 'action'
         # When an action event is processed, store the event.
         @redis.hset(event.id + ':actions', timestamp, event.state)
-        @redis.hincrby('event_counters', 'action', 1) if event.ok?
+        @redis.hincrby('event_counters', 'action', 1)
+        @redis.hincrby("event_counters:#{@instance_id}", 'action', 1)
 
         if event.acknowledgement? && event.acknowledgement_id
           @redis.hdel('unacknowledged_failures', event.acknowledgement_id)
         end
       when 'shutdown'