#!/usr/bin/env ruby require 'yajl/json_gem' require 'flapjack/patches' require 'flapjack/data/contact' require 'flapjack/data/event' require 'flapjack/data/entity' # TODO might want to split the class methods out to a separate class, DAO pattern # ( http://en.wikipedia.org/wiki/Data_access_object ). module Flapjack module Data class EntityCheck STATE_OK = 'ok' STATE_WARNING = 'warning' STATE_CRITICAL = 'critical' STATE_UNKNOWN = 'unknown' attr_accessor :entity, :check # TODO probably shouldn't always be creating on query -- work out when this should be happening def self.for_event_id(event_id, options = {}) raise "Redis connection not set" unless redis = options[:redis] entity_name, check = event_id.split(':') self.new(Flapjack::Data::Entity.find_by_name(entity_name, :redis => redis, :create => true), check, :redis => redis) end # TODO probably shouldn't always be creating on query -- work out when this should be happening def self.for_entity_name(entity_name, check, options = {}) raise "Redis connection not set" unless redis = options[:redis] self.new(Flapjack::Data::Entity.find_by_name(entity_name, :redis => redis, :create => true), check, :redis => redis) end def self.for_entity_id(entity_id, check, options = {}) raise "Redis connection not set" unless redis = options[:redis] self.new(Flapjack::Data::Entity.find_by_id(entity_id, :redis => redis), check, :redis => redis) end def self.for_entity(entity, check, options = {}) raise "Redis connection not set" unless redis = options[:redis] self.new(entity, check, :redis => redis) end def entity_name entity.name end # takes a key "entity:check", returns true if the check is in unscheduled # maintenance def in_unscheduled_maintenance? @redis.exists("#{@key}:unscheduled_maintenance") end # returns true if the check is in scheduled maintenance def in_scheduled_maintenance? @redis.exists("#{@key}:scheduled_maintenance") end # return data about current maintenance (scheduled or unscheduled, as specified) def current_maintenance(opts) sched = opts[:scheduled] ? 'scheduled' : 'unscheduled' ts = @redis.get("#{@key}:#{sched}_maintenance") return unless ts {:start_time => ts.to_i, :duration => @redis.zscore("#{@key}:#{sched}_maintenances", ts), :summary => @redis.get("#{@key}:#{ts}:#{sched}_maintenance:summary"), } end # TODO move to Event def create_acknowledgement(options = {}) event = { 'type' => 'action', 'state' => 'acknowledgement', 'summary' => options['summary'], 'duration' => options['duration'], 'acknowledgement_id' => options['acknowledgement_id'], 'entity' => entity.name, 'check' => check } Flapjack::Data::Event.add(event, :redis => @redis) end # TODO move to Event def test_notifications(options = {}) event = { 'type' => 'action', 'state' => 'test_notifications', 'summary' => options['summary'], 'entity' => entity.name, 'check' => check } Flapjack::Data::Event.add(event, :redis => @redis) end # FIXME: need to add summary to summary of existing unscheduled maintenance if there is # one, and extend duration / expiry time, instead of creating a separate unscheduled # outage as we are doing now... def create_unscheduled_maintenance(opts = {}) start_time = opts[:start_time] # unix timestamp duration = opts[:duration] # seconds summary = opts[:summary] time_remaining = (start_time + duration) - Time.now.to_i if time_remaining > 0 @redis.setex("#{@key}:unscheduled_maintenance", time_remaining, start_time) end @redis.zadd("#{@key}:unscheduled_maintenances", duration, start_time) @redis.set("#{@key}:#{start_time}:unscheduled_maintenance:summary", summary) @redis.zadd("#{@key}:sorted_unscheduled_maintenance_timestamps", start_time, start_time) end # ends any unscheduled maintenance def end_unscheduled_maintenance(opts = {}) defaults = { :end_time => Time.now.to_i } options = defaults.merge(opts) end_time = options[:end_time] if (um_start = @redis.get("#{@key}:unscheduled_maintenance")) duration = end_time - um_start.to_i @logger.debug("ending unscheduled downtime for #{@key} at #{Time.at(end_time).to_s}") if @logger @redis.del("#{@key}:unscheduled_maintenance") @redis.zadd("#{@key}:unscheduled_maintenances", duration, um_start) @redis.zadd("#{@key}:sorted_unscheduled_maintenance_timestamps", um_start, um_start) else @logger.debug("end_unscheduled_maintenance called for #{@key} but none found") if @logger end end # creates a scheduled maintenance period for a check # TODO: consider adding some validation to the data we're adding in here # eg start_time is a believable unix timestamp (not in the past and not too # far in the future), duration is within some bounds... def create_scheduled_maintenance(opts = {}) start_time = opts[:start_time] # unix timestamp duration = opts[:duration] # seconds summary = opts[:summary] @redis.zadd("#{@key}:scheduled_maintenances", duration, start_time) @redis.set("#{@key}:#{start_time}:scheduled_maintenance:summary", summary) @redis.zadd("#{@key}:sorted_scheduled_maintenance_timestamps", start_time, start_time) # scheduled maintenance periods have changed, revalidate update_current_scheduled_maintenance(:revalidate => true) end # change the end time of a scheduled maintenance (including when one is current) # TODO allow to update summary as well def update_scheduled_maintenance(start_time, patches = {}) # check if there is such a scheduled maintenance period old_duration = @redis.zscore("#{@key}:scheduled_maintenances", start_time) raise ArgumentError, 'no such scheduled maintenance period can be found' unless old_duration raise ArgumentError, 'no handled patches have been supplied' unless patches[:end_time] if patches[:end_time] end_time = patches[:end_time] raise ArgumentError, "end time must be after start time" unless end_time > start_time old_end_time = start_time + old_duration duration = end_time - start_time @redis.zadd("#{@key}:scheduled_maintenances", duration, start_time) end # scheduled maintenance periods have changed, revalidate update_current_scheduled_maintenance(:revalidate => true) end # delete a scheduled maintenance def delete_scheduled_maintenance(opts = {}) start_time = opts[:start_time] @redis.del("#{@key}:#{start_time}:scheduled_maintenance:summary") @redis.zrem("#{@key}:scheduled_maintenances", start_time) @redis.zremrangebyscore("#{@key}:sorted_scheduled_maintenance_timestamps", start_time, start_time) # scheduled maintenance periods have changed, revalidate update_current_scheduled_maintenance(:revalidate => true) end # if not in scheduled maintenance, looks in scheduled maintenance list for a check to see if # current state should be set to scheduled maintenance, and sets it as appropriate def update_current_scheduled_maintenance(opts = {}) if opts[:revalidate] @redis.del("#{@key}:scheduled_maintenance") else return if in_scheduled_maintenance? end # are we within a scheduled maintenance period? t = Time.now.to_i current_sched_ms = maintenances(nil, nil, :scheduled => true).select {|sm| (sm[:start_time] <= t) && (t < sm[:end_time]) } return if current_sched_ms.empty? # yes! so set current scheduled maintenance # if multiple scheduled maintenances found, find the end_time furthest in the future most_futuristic = current_sched_ms.max {|sm| sm[:end_time] } start_time = most_futuristic[:start_time] duration = most_futuristic[:duration] @redis.setex("#{@key}:scheduled_maintenance", duration.to_i, start_time) end # returns nil if no previous state; this must be considered as a possible # state by classes using this model def state @redis.hget("check:#{@key}", 'state') end def update_state(state, options = {}) return unless validate_state(state) timestamp = options[:timestamp] || Time.now.to_i client = options[:client] summary = options[:summary] count = options[:count] # Note the current state (for speedy lookups) @redis.hset("check:#{@key}", 'state', state) # FIXME: rename to last_state_change? @redis.hset("check:#{@key}", 'last_change', timestamp) # Retain all state changes for entity:check pair @redis.rpush("#{@key}:states", timestamp) @redis.set("#{@key}:#{timestamp}:state", state) @redis.set("#{@key}:#{timestamp}:summary", summary) if summary @redis.set("#{@key}:#{timestamp}:count", count) if count @redis.zadd("#{@key}:sorted_state_timestamps", timestamp, timestamp) case state when STATE_WARNING, STATE_CRITICAL, STATE_UNKNOWN @redis.zadd('failed_checks', timestamp, @key) # FIXME: Iterate through a list of tags associated with an entity:check pair, and update counters @redis.zadd("failed_checks:client:#{client}", timestamp, @key) if client else @redis.zrem("failed_checks", @key) # FIXME: Iterate through a list of tags associated with an entity:check pair, and update counters @redis.zrem("failed_checks:client:#{client}", @key) if client end end def last_update lu = @redis.hget("check:#{@key}", 'last_update') return unless (lu && lu =~ /^\d+$/) lu.to_i end def last_update=(timestamp) @redis.hset("check:#{@key}", 'last_update', timestamp) end def last_change lc = @redis.hget("check:#{@key}", 'last_change') return unless (lc && lc =~ /^\d+$/) lc.to_i end def last_problem_notification lpn = @redis.get("#{@key}:last_problem_notification") return unless (lpn && lpn =~ /^\d+$/) lpn.to_i end def last_recovery_notification lrn = @redis.get("#{@key}:last_recovery_notification") return unless (lrn && lrn =~ /^\d+$/) lrn.to_i end def last_acknowledgement_notification lan = @redis.get("#{@key}:last_acknowledgement_notification") return unless (lan && lan =~ /^\d+$/) lan.to_i end def last_notifications_of_each_type ln = {:problem => last_problem_notification, :recovery => last_recovery_notification, :acknowledgement => last_acknowledgement_notification } ln end # unpredictable results if there are multiple notifications of different # types sent at the same time def last_notification nils = { :type => nil, :timestamp => nil } lne = last_notifications_of_each_type ln = lne.delete_if {|type, timestamp| timestamp.nil? || timestamp.to_i == 0 } return nils unless ln.length > 0 lns = ln.sort_by { |type, timestamp| timestamp }.last { :type => lns[0], :timestamp => lns[1] } end def event_count_at(timestamp) eca = @redis.get("#{@key}:#{timestamp}:count") return unless (eca && eca =~ /^\d+$/) eca.to_i end def failed? [STATE_WARNING, STATE_CRITICAL, STATE_UNKNOWN].include?( state ) end def ok? [STATE_OK].include?( state ) end def summary timestamp = @redis.lindex("#{@key}:states", -1) @redis.get("#{@key}:#{timestamp}:summary") end # Returns a list of states for this entity check, sorted by timestamp. # # start_time and end_time should be passed as integer timestamps; these timestamps # will be considered inclusively, so, e.g. coverage for a day should go # from midnight to 11:59:59 PM. Pass nil for either end to leave that # side unbounded. def historical_states(start_time, end_time, opts = {}) start_time ||= '-inf' end_time ||= '+inf' order = opts[:order] query = (order && 'desc'.eql?(order.downcase)) ? :zrevrangebyscore : :zrangebyscore state_ts = @redis.send(query, "#{@key}:sorted_state_timestamps", start_time, end_time) state_data = nil @redis.multi do |r| state_data = state_ts.collect {|ts| {:timestamp => ts.to_i, :state => r.get("#{@key}:#{ts}:state"), :summary => r.get("#{@key}:#{ts}:summary")} } end # The redis commands in a pipeline block return future objects, which # must be evaluated. This relies on a patch in flapjack/patches.rb to # make the Future objects report their class. state_data.collect {|sd| sd.merge!(sd) {|k,ov,nv| (nv.class == Redis::Future) ? nv.value : nv } } end # requires a known state timestamp, i.e. probably one returned via # historical_states. will find the one before that in the sorted set, # if any. def historical_state_before(timestamp) pos = @redis.zrank("#{@key}:sorted_state_timestamps", timestamp) return if pos < 1 ts = @redis.zrange("#{@key}:sorted_state_timestamps", pos - 1, pos) return if ts.nil? || ts.empty? {:timestamp => ts.first.to_i, :state => @redis.get("#{@key}:#{ts.first}:state"), :summary => @redis.get("#{@key}:#{ts.first}:summary")} end # Returns a list of maintenance periods (either unscheduled or scheduled) for this # entity check, sorted by timestamp. # # start_time and end_time should be passed as integer timestamps; these timestamps # will be considered inclusively, so, e.g. coverage for a day should go # from midnight to 11:59:59 PM. Pass nil for either end to leave that # side unbounded. def maintenances(start_time, end_time, opts = {}) sched = opts[:scheduled] ? 'scheduled' : 'unscheduled' start_time ||= '-inf' end_time ||= '+inf' order = opts[:order] query = (order && 'desc'.eql?(order.downcase)) ? :zrevrangebyscore : :zrangebyscore maint_ts = @redis.send(query, "#{@key}:sorted_#{sched}_maintenance_timestamps", start_time, end_time) maint_data = nil @redis.multi do |r| maint_data = maint_ts.collect {|ts| {:start_time => ts.to_i, :duration => r.zscore("#{@key}:#{sched}_maintenances", ts), :summary => r.get("#{@key}:#{ts}:#{sched}_maintenance:summary"), } } end # The redis commands in a pipeline block return future objects, which # must be evaluated. This relies on a patch in flapjack/patches.rb to # make the Future objects report their class. maint_data.collect {|md| md.merge!(md) {|k,ov,nv| (nv.class == Redis::Future) ? nv.value : nv } md[:end_time] = (md[:start_time] + md[:duration]).floor md } end # takes a check, looks up contacts that are interested in this check (or in the check's entity) # and returns an array of contact records def contacts contact_ids = @redis.smembers("contacts_for:#{entity.id}:#{check}") if @logger @logger.debug("#{contact_ids.length} contact(s) for #{entity.id}:#{check}: " + contact_ids.inspect) end entity.contacts + contact_ids.collect {|c_id| Flapjack::Data::Contact.find_by_id(c_id, :redis => @redis) }.compact end private def initialize(ent, che, options = {}) raise "Redis connection not set" unless @redis = options[:redis] raise "Invalid entity" unless @entity = ent raise "Invalid check" unless @check = che @key = "#{ent.name}:#{che}" end def validate_state(state) [STATE_OK, STATE_WARNING, STATE_CRITICAL, STATE_UNKNOWN].include?(state) end end end end