Class: GitLab::Exporter::SidekiqProber

Inherits:
Object
  • Object
show all
Defined in:
lib/gitlab_exporter/sidekiq.rb

Overview

A prober for Sidekiq queues

It takes the Redis URL Sidekiq is connected to

Constant Summary collapse

PROBE_JOBS_LIMIT =

The maximum depth (from the head) of each queue to probe. Probing the entirety of a very large queue will take longer and run the risk of timing out. But when we have a very large queue, we are most in need of reliable metrics. This trades off completeness for predictability by only taking a limited amount of items from the head of the queue.

1_000
POOL_SIZE = 
3
POOL_TIMEOUT =

This timeout is configured to higher interval than scrapping of Prometheus to ensure that connection is kept instead of needed to be re-initialized

90
SIDEKIQ_REDIS_LOCK =

Lock for Sidekiq.redis which we need to modify, but is not concurrency safe.

Mutex.new

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(metrics: PrometheusMetrics.new, logger: nil, **opts) ⇒ SidekiqProber

Returns a new instance of SidekiqProber.



39
40
41
42
43
 
# File 'lib/gitlab_exporter/sidekiq.rb', line 39

def initialize(metrics: PrometheusMetrics.new, logger: nil, **opts)
  @opts    = opts
  @metrics = metrics
  @logger  = logger
end

Class Method Details

.connection_pool ⇒ Object 



31
32
33
34
35
36
37
 
# File 'lib/gitlab_exporter/sidekiq.rb', line 31

def self.connection_pool
  @@connection_pool ||= Hash.new do |h, connection_hash| # rubocop:disable Style/ClassVars
    config = connection_hash.merge(pool_timeout: POOL_TIMEOUT, size: POOL_SIZE)

    h[connection_hash] = Sidekiq::RedisConnection.create(config)
  end
end

Instance Method Details

#probe_dead ⇒ Object 



166
167
168
169
170
171
172
173
174
175
 
# File 'lib/gitlab_exporter/sidekiq.rb', line 166

def probe_dead
  puts "[DEPRECATED] probe_dead is now considered obsolete and will be removed in future major versions,"\
       " please use probe_stats instead"

  with_sidekiq do
    @metrics.add("sidekiq_dead_jobs", Sidekiq::Stats.new.dead_size)
  end

  self
end

#probe_future_sets ⇒ Object 



83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
 
# File 'lib/gitlab_exporter/sidekiq.rb', line 83

def probe_future_sets
  now = Time.now.to_f
  with_sidekiq do
    Sidekiq.redis do |conn|
      Sidekiq::Scheduled::SETS.each do |set|
        # Default to 0; if all jobs are due in the future, there is no "negative" delay.
        delay = 0

        _job, timestamp = conn.zrangebyscore(set, "-inf", now.to_s, limit: [0, 1], withscores: true).first
        delay = now - timestamp if timestamp

        @metrics.add("sidekiq_#{set}_set_processing_delay_seconds", delay)

        # zcount is O(log(N)) (prob. binary search), so is still quick even with large sets
        @metrics.add("sidekiq_#{set}_set_backlog_count",
                     conn.zcount(set, "-inf", now.to_s))
      end
    end
  end
end

#probe_jobs ⇒ Object 



76
77
78
79
80
81
 
# File 'lib/gitlab_exporter/sidekiq.rb', line 76

def probe_jobs
  puts "[REMOVED] probe_jobs is now considered obsolete and does not emit any metrics,"\
       " please use probe_jobs_limit instead"

  self
end

#probe_jobs_limit ⇒ Object

Count worker classes present in Sidekiq queues. This only looks at the first PROBE_JOBS_LIMIT jobs in each queue. This means that we run a single LRANGE command for each queue, which does not block other commands. For queues over PROBE_JOBS_LIMIT in size, this means that we will not have completely accurate statistics, but the probe performance will also not degrade as the queue gets larger.



110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
 
# File 'lib/gitlab_exporter/sidekiq.rb', line 110

def probe_jobs_limit
  with_sidekiq do
    job_stats = Hash.new(0)

    Sidekiq::Queue.all.each do |queue|
      Sidekiq.redis do |conn|
        conn.lrange("queue:#{queue.name}", 0, PROBE_JOBS_LIMIT).each do |job|
          job_class = Sidekiq.load_json(job)["class"]

          job_stats[job_class] += 1
        end
      end
    end

    job_stats.each do |class_name, count|
      @metrics.add("sidekiq_enqueued_jobs", count, name: class_name)
    end
  end

  self
end

#probe_queues ⇒ Object 



64
65
66
67
68
69
70
71
72
73
74
 
# File 'lib/gitlab_exporter/sidekiq.rb', line 64

def probe_queues
  with_sidekiq do
    Sidekiq::Queue.all.each do |queue|
      @metrics.add("sidekiq_queue_size", queue.size, name: queue.name)
      @metrics.add("sidekiq_queue_latency_seconds", queue.latency, name: queue.name)
      @metrics.add("sidekiq_queue_paused", queue.paused? ? 1 : 0, name: queue.name)
    end
  end

  self
end

#probe_retries ⇒ Object 



150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
 
# File 'lib/gitlab_exporter/sidekiq.rb', line 150

def probe_retries
  with_sidekiq do
    retry_stats = Hash.new(0)

    Sidekiq::RetrySet.new.map do |job|
      retry_stats[job.klass] += 1
    end

    retry_stats.each do |class_name, count|
      @metrics.add("sidekiq_to_be_retried_jobs", count, name: class_name)
    end
  end

  self
end

#probe_stats ⇒ Object 



45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
 
# File 'lib/gitlab_exporter/sidekiq.rb', line 45

def probe_stats
  with_sidekiq do
    stats = Sidekiq::Stats.new

    @metrics.add("sidekiq_jobs_processed_total", stats.processed)
    @metrics.add("sidekiq_jobs_failed_total", stats.failed)
    @metrics.add("sidekiq_jobs_enqueued_size", stats.enqueued)
    @metrics.add("sidekiq_jobs_scheduled_size", stats.scheduled_size)
    @metrics.add("sidekiq_jobs_retry_size", stats.retry_size)
    @metrics.add("sidekiq_jobs_dead_size", stats.dead_size)

    @metrics.add("sidekiq_default_queue_latency_seconds", stats.default_queue_latency)
    @metrics.add("sidekiq_processes_size", stats.processes_size)
    @metrics.add("sidekiq_workers_size", stats.workers_size)
  end

  self
end

#probe_workers ⇒ Object 



132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
 
# File 'lib/gitlab_exporter/sidekiq.rb', line 132

def probe_workers
  with_sidekiq do
    worker_stats = Hash.new(0)

    Sidekiq::Workers.new.map do |_pid, _tid, work|
      job_klass = work["payload"]["class"]

      worker_stats[job_klass] += 1
    end

    worker_stats.each do |class_name, count|
      @metrics.add("sidekiq_running_jobs", count, name: class_name)
    end
  end

  self
end

#write_to(target) ⇒ Object 



177
178
179
 
# File 'lib/gitlab_exporter/sidekiq.rb', line 177

def write_to(target)
  target.write(@metrics.to_s)
end