Class: Sidekiq::Launcher
- Inherits:
-
Object
- Object
- Sidekiq::Launcher
- Includes:
- Util
- Defined in:
- lib/sidekiq/launcher.rb
Overview
The Launcher starts the Manager and Poller threads and provides the process heartbeat.
Constant Summary collapse
- STATS_TTL =
5 years
5 * 365 * 24 * 60 * 60
- PROCTITLES =
[ proc { "sidekiq" }, proc { Sidekiq::VERSION }, proc { |me, data| data["tag"] }, proc { |me, data| "[#{Processor::WORKER_STATE.size} of #{data["concurrency"]} busy]" }, proc { |me, data| "stopping" if me.stopping? } ]
- BEAT_PAUSE =
5- RTT_READINGS =
We run the heartbeat every five seconds. Capture five samples of RTT, log a warning if each sample is above our warning threshold.
RingBuffer.new(5)
- RTT_WARNING_LEVEL =
50_000- MEMORY_GRABBER =
case RUBY_PLATFORM when /linux/ ->(pid) { IO.readlines("/proc/#{$$}/status").each do |line| next unless line.start_with?("VmRSS:") break line.split[1].to_i end } when /darwin|bsd/ ->(pid) { `ps -o pid,rss -p #{pid}`.lines.last.split.last.to_i } else ->(pid) { 0 } end
Constants included from Util
Instance Attribute Summary collapse
-
#fetcher ⇒ Object
Returns the value of attribute fetcher.
-
#manager ⇒ Object
Returns the value of attribute manager.
-
#poller ⇒ Object
Returns the value of attribute poller.
Class Method Summary collapse
Instance Method Summary collapse
- #check_rtt ⇒ Object
- #clear_heartbeat ⇒ Object
- #heartbeat ⇒ Object
-
#initialize(options) ⇒ Launcher
constructor
A new instance of Launcher.
- #memory_usage(pid) ⇒ Object
-
#quiet ⇒ Object
Stops this instance from processing any more jobs,.
- #run ⇒ Object
- #start_heartbeat ⇒ Object
-
#stop ⇒ Object
Shuts down the process.
- #stopping? ⇒ Boolean
- #to_data ⇒ Object
- #to_json ⇒ Object
- #❤ ⇒ Object
Methods included from Util
#fire_event, #hostname, #identity, #logger, #process_nonce, #redis, #safe_thread, #tid, #wait_for, #watchdog
Methods included from ExceptionHandler
Constructor Details
Instance Attribute Details
#fetcher ⇒ Object
Returns the value of attribute fetcher.
22 23 24 |
# File 'lib/sidekiq/launcher.rb', line 22 def fetcher @fetcher end |
#manager ⇒ Object
Returns the value of attribute manager.
22 23 24 |
# File 'lib/sidekiq/launcher.rb', line 22 def manager @manager end |
#poller ⇒ Object
Returns the value of attribute poller.
22 23 24 |
# File 'lib/sidekiq/launcher.rb', line 22 def poller @poller end |
Class Method Details
.flush_stats ⇒ Object
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 |
# File 'lib/sidekiq/launcher.rb', line 102 def self.flush_stats fails = Processor::FAILURE.reset procd = Processor::PROCESSED.reset return if fails + procd == 0 nowdate = Time.now.utc.strftime("%Y-%m-%d") begin Sidekiq.redis do |conn| conn.pipelined do conn.incrby("stat:processed", procd) conn.incrby("stat:processed:#{nowdate}", procd) conn.expire("stat:processed:#{nowdate}", STATS_TTL) conn.incrby("stat:failed", fails) conn.incrby("stat:failed:#{nowdate}", fails) conn.expire("stat:failed:#{nowdate}", STATS_TTL) end end rescue => ex # we're exiting the process, things might be shut down so don't # try to handle the exception Sidekiq.logger.warn("Unable to flush stats: #{ex}") end end |
Instance Method Details
#check_rtt ⇒ Object
199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 |
# File 'lib/sidekiq/launcher.rb', line 199 def check_rtt a = b = 0 Sidekiq.redis do |x| a = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC, :microsecond) x.ping b = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC, :microsecond) end rtt = b - a RTT_READINGS << rtt # Ideal RTT for Redis is < 1000µs # Workable is < 10,000µs # Log a warning if it's a disaster. if RTT_READINGS.all? { |x| x > RTT_WARNING_LEVEL } Sidekiq.logger.warn <<~EOM Your Redis network connection is performing extremely poorly. Last RTT readings were #{RTT_READINGS.buffer.inspect}, ideally these should be < 1000. Ensure Redis is running in the same AZ or datacenter as Sidekiq. If these values are close to 100,000, that means your Sidekiq process may be CPU overloaded; see https://github.com/mperham/sidekiq/discussions/5039 EOM RTT_READINGS.reset end rtt end |
#clear_heartbeat ⇒ Object
82 83 84 85 86 87 88 89 90 91 92 93 94 |
# File 'lib/sidekiq/launcher.rb', line 82 def clear_heartbeat # Remove record from Redis since we are shutting down. # Note we don't stop the heartbeat thread; if the process # doesn't actually exit, it'll reappear in the Web UI. Sidekiq.redis do |conn| conn.pipelined do conn.srem("processes", identity) conn.unlink("#{identity}:workers") end end rescue # best effort, ignore network errors end |
#heartbeat ⇒ Object
96 97 98 99 100 |
# File 'lib/sidekiq/launcher.rb', line 96 def heartbeat $0 = PROCTITLES.map { |proc| proc.call(self, to_data) }.compact.join(" ") ❤ end |
#memory_usage(pid) ⇒ Object
240 241 242 |
# File 'lib/sidekiq/launcher.rb', line 240 def memory_usage(pid) MEMORY_GRABBER.call(pid) end |
#quiet ⇒ Object
Stops this instance from processing any more jobs,
40 41 42 43 44 |
# File 'lib/sidekiq/launcher.rb', line 40 def quiet @done = true @manager.quiet @poller.terminate end |
#run ⇒ Object
32 33 34 35 36 |
# File 'lib/sidekiq/launcher.rb', line 32 def run @thread = safe_thread("heartbeat", &method(:start_heartbeat)) @poller.start @manager.start end |
#start_heartbeat ⇒ Object
74 75 76 77 78 79 80 |
# File 'lib/sidekiq/launcher.rb', line 74 def start_heartbeat loop do heartbeat sleep BEAT_PAUSE end Sidekiq.logger.info("Heartbeat stopping...") end |
#stop ⇒ Object
Shuts down the process. This method does not return until all work is complete and cleaned up. It can take up to the timeout to complete.
49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 |
# File 'lib/sidekiq/launcher.rb', line 49 def stop deadline = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC) + @options[:timeout] @done = true @manager.quiet @poller.terminate @manager.stop(deadline) # Requeue everything in case there was a worker who grabbed work while stopped # This call is a no-op in Sidekiq but necessary for Sidekiq Pro. strategy = @options[:fetch] strategy.bulk_requeue([], @options) clear_heartbeat end |
#stopping? ⇒ Boolean
66 67 68 |
# File 'lib/sidekiq/launcher.rb', line 66 def stopping? @done end |
#to_data ⇒ Object
244 245 246 247 248 249 250 251 252 253 254 255 |
# File 'lib/sidekiq/launcher.rb', line 244 def to_data @data ||= { "hostname" => hostname, "started_at" => Time.now.to_f, "pid" => ::Process.pid, "tag" => @options[:tag] || "", "concurrency" => @options[:concurrency], "queues" => @options[:queues].uniq, "labels" => @options[:labels], "identity" => identity } end |
#to_json ⇒ Object
257 258 259 260 261 |
# File 'lib/sidekiq/launcher.rb', line 257 def to_json # this data changes infrequently so dump it to a string # now so we don't need to dump it every heartbeat. @json ||= Sidekiq.dump_json(to_data) end |
#❤ ⇒ Object
128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 |
# File 'lib/sidekiq/launcher.rb', line 128 def ❤ key = identity fails = procd = 0 begin fails = Processor::FAILURE.reset procd = Processor::PROCESSED.reset curstate = Processor::WORKER_STATE.dup workers_key = "#{key}:workers" nowdate = Time.now.utc.strftime("%Y-%m-%d") Sidekiq.redis do |conn| conn.multi do conn.incrby("stat:processed", procd) conn.incrby("stat:processed:#{nowdate}", procd) conn.expire("stat:processed:#{nowdate}", STATS_TTL) conn.incrby("stat:failed", fails) conn.incrby("stat:failed:#{nowdate}", fails) conn.expire("stat:failed:#{nowdate}", STATS_TTL) conn.unlink(workers_key) curstate.each_pair do |tid, hash| conn.hset(workers_key, tid, Sidekiq.dump_json(hash)) end conn.expire(workers_key, 60) end end rtt = check_rtt fails = procd = 0 kb = memory_usage(::Process.pid) _, exists, _, _, msg = Sidekiq.redis { |conn| conn.multi { conn.sadd("processes", key) conn.exists?(key) conn.hmset(key, "info", to_json, "busy", curstate.size, "beat", Time.now.to_f, "rtt_us", rtt, "quiet", @done, "rss", kb) conn.expire(key, 60) conn.rpop("#{key}-signals") } } # first heartbeat or recovering from an outage and need to reestablish our heartbeat fire_event(:heartbeat) unless exists return unless msg ::Process.kill(msg, ::Process.pid) rescue => e # ignore all redis/network issues logger.error("heartbeat: #{e}") # don't lose the counts if there was a network issue Processor::PROCESSED.incr(procd) Processor::FAILURE.incr(fails) end end |