prometheus_exporter 0.5.3 → 0.8.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +4 -4
- data/.github/workflows/ci.yml +42 -0
- data/.gitignore +2 -0
- data/.rubocop.yml +5 -0
- data/Appraisals +10 -0
- data/CHANGELOG +28 -0
- data/README.md +138 -16
- data/bin/prometheus_exporter +29 -2
- data/gemfiles/.bundle/config +2 -0
- data/gemfiles/ar_60.gemfile +5 -0
- data/gemfiles/ar_61.gemfile +7 -0
- data/lib/prometheus_exporter.rb +1 -0
- data/lib/prometheus_exporter/client.rb +39 -8
- data/lib/prometheus_exporter/instrumentation.rb +1 -0
- data/lib/prometheus_exporter/instrumentation/active_record.rb +19 -12
- data/lib/prometheus_exporter/instrumentation/delayed_job.rb +3 -2
- data/lib/prometheus_exporter/instrumentation/method_profiler.rb +2 -1
- data/lib/prometheus_exporter/instrumentation/process.rb +1 -1
- data/lib/prometheus_exporter/instrumentation/puma.rb +17 -5
- data/lib/prometheus_exporter/instrumentation/resque.rb +40 -0
- data/lib/prometheus_exporter/instrumentation/sidekiq.rb +44 -3
- data/lib/prometheus_exporter/instrumentation/sidekiq_queue.rb +13 -2
- data/lib/prometheus_exporter/instrumentation/unicorn.rb +1 -1
- data/lib/prometheus_exporter/middleware.rb +40 -17
- data/lib/prometheus_exporter/server.rb +1 -0
- data/lib/prometheus_exporter/server/active_record_collector.rb +2 -1
- data/lib/prometheus_exporter/server/collector.rb +1 -0
- data/lib/prometheus_exporter/server/delayed_job_collector.rb +9 -8
- data/lib/prometheus_exporter/server/puma_collector.rb +9 -1
- data/lib/prometheus_exporter/server/resque_collector.rb +54 -0
- data/lib/prometheus_exporter/server/runner.rb +13 -3
- data/lib/prometheus_exporter/server/sidekiq_collector.rb +2 -2
- data/lib/prometheus_exporter/server/web_collector.rb +2 -5
- data/lib/prometheus_exporter/server/web_server.rb +33 -23
- data/lib/prometheus_exporter/version.rb +1 -1
- data/prometheus_exporter.gemspec +7 -3
- metadata +59 -11
- data/.travis.yml +0 -12
data/lib/prometheus_exporter.rb
CHANGED
@@ -2,6 +2,7 @@
|
|
2
2
|
|
3
3
|
require 'socket'
|
4
4
|
require 'thread'
|
5
|
+
require 'logger'
|
5
6
|
|
6
7
|
module PrometheusExporter
|
7
8
|
class Client
|
@@ -53,24 +54,32 @@ module PrometheusExporter
|
|
53
54
|
MAX_SOCKET_AGE = 25
|
54
55
|
MAX_QUEUE_SIZE = 10_000
|
55
56
|
|
57
|
+
attr_reader :logger
|
58
|
+
|
56
59
|
def initialize(
|
57
60
|
host: ENV.fetch('PROMETHEUS_EXPORTER_HOST', 'localhost'),
|
58
61
|
port: ENV.fetch('PROMETHEUS_EXPORTER_PORT', PrometheusExporter::DEFAULT_PORT),
|
59
62
|
max_queue_size: nil,
|
60
63
|
thread_sleep: 0.5,
|
61
64
|
json_serializer: nil,
|
62
|
-
custom_labels: nil
|
65
|
+
custom_labels: nil,
|
66
|
+
logger: Logger.new(STDERR),
|
67
|
+
log_level: Logger::WARN
|
63
68
|
)
|
69
|
+
@logger = logger
|
70
|
+
@logger.level = log_level
|
64
71
|
@metrics = []
|
65
72
|
|
66
73
|
@queue = Queue.new
|
74
|
+
|
67
75
|
@socket = nil
|
68
76
|
@socket_started = nil
|
77
|
+
@socket_pid = nil
|
69
78
|
|
70
79
|
max_queue_size ||= MAX_QUEUE_SIZE
|
71
80
|
max_queue_size = max_queue_size.to_i
|
72
81
|
|
73
|
-
if max_queue_size
|
82
|
+
if max_queue_size <= 0
|
74
83
|
raise ArgumentError, "max_queue_size must be larger than 0"
|
75
84
|
end
|
76
85
|
|
@@ -107,14 +116,23 @@ module PrometheusExporter
|
|
107
116
|
end
|
108
117
|
|
109
118
|
def send_json(obj)
|
110
|
-
payload =
|
119
|
+
payload =
|
120
|
+
if @custom_labels
|
121
|
+
if obj[:custom_labels]
|
122
|
+
obj.merge(custom_labels: @custom_labels.merge(obj[:custom_labels]))
|
123
|
+
else
|
124
|
+
obj.merge(custom_labels: @custom_labels)
|
125
|
+
end
|
126
|
+
else
|
127
|
+
obj
|
128
|
+
end
|
111
129
|
send(@json_serializer.dump(payload))
|
112
130
|
end
|
113
131
|
|
114
132
|
def send(str)
|
115
133
|
@queue << str
|
116
134
|
if @queue.length > @max_queue_size
|
117
|
-
|
135
|
+
logger.warn "Prometheus Exporter client is dropping message cause queue is full"
|
118
136
|
@queue.pop
|
119
137
|
end
|
120
138
|
|
@@ -132,7 +150,7 @@ module PrometheusExporter
|
|
132
150
|
@socket.write(message)
|
133
151
|
@socket.write("\r\n")
|
134
152
|
rescue => e
|
135
|
-
|
153
|
+
logger.warn "Prometheus Exporter is dropping a message: #{e}"
|
136
154
|
@socket = nil
|
137
155
|
raise
|
138
156
|
end
|
@@ -157,7 +175,7 @@ module PrometheusExporter
|
|
157
175
|
close_socket_if_old!
|
158
176
|
process_queue
|
159
177
|
rescue => e
|
160
|
-
|
178
|
+
logger.error "Prometheus Exporter, failed to send message #{e}"
|
161
179
|
end
|
162
180
|
|
163
181
|
def ensure_worker_thread!
|
@@ -173,11 +191,14 @@ module PrometheusExporter
|
|
173
191
|
end
|
174
192
|
end
|
175
193
|
end
|
194
|
+
rescue ThreadError => e
|
195
|
+
raise unless e.message =~ /can't alloc thread/
|
196
|
+
logger.error "Prometheus Exporter, failed to send message ThreadError #{e}"
|
176
197
|
end
|
177
198
|
|
178
199
|
def close_socket!
|
179
200
|
begin
|
180
|
-
if @socket
|
201
|
+
if @socket && !@socket.closed?
|
181
202
|
@socket.write("0\r\n")
|
182
203
|
@socket.write("\r\n")
|
183
204
|
@socket.flush
|
@@ -191,12 +212,20 @@ module PrometheusExporter
|
|
191
212
|
end
|
192
213
|
|
193
214
|
def close_socket_if_old!
|
194
|
-
if @socket && ((@socket_started + MAX_SOCKET_AGE) < Time.now.to_f)
|
215
|
+
if @socket_pid == Process.pid && @socket && @socket_started && ((@socket_started + MAX_SOCKET_AGE) < Time.now.to_f)
|
195
216
|
close_socket!
|
196
217
|
end
|
197
218
|
end
|
198
219
|
|
199
220
|
def ensure_socket!
|
221
|
+
# if process was forked socket may be owned by parent
|
222
|
+
# leave it alone and reset
|
223
|
+
if @socket_pid != Process.pid
|
224
|
+
@socket = nil
|
225
|
+
@socket_started = nil
|
226
|
+
@socket_pid = nil
|
227
|
+
end
|
228
|
+
|
200
229
|
close_socket_if_old!
|
201
230
|
if !@socket
|
202
231
|
@socket = TCPSocket.new @host, @port
|
@@ -207,12 +236,14 @@ module PrometheusExporter
|
|
207
236
|
@socket.write("Content-Type: application/octet-stream\r\n")
|
208
237
|
@socket.write("\r\n")
|
209
238
|
@socket_started = Time.now.to_f
|
239
|
+
@socket_pid = Process.pid
|
210
240
|
end
|
211
241
|
|
212
242
|
nil
|
213
243
|
rescue
|
214
244
|
@socket = nil
|
215
245
|
@socket_started = nil
|
246
|
+
@socket_pid = nil
|
216
247
|
raise
|
217
248
|
end
|
218
249
|
|
@@ -7,9 +7,11 @@ module PrometheusExporter::Instrumentation
|
|
7
7
|
|
8
8
|
def self.start(client: nil, frequency: 30, custom_labels: {}, config_labels: [])
|
9
9
|
|
10
|
-
|
10
|
+
client ||= PrometheusExporter::Client.default
|
11
|
+
|
12
|
+
# Not all rails versions support connection pool stats
|
11
13
|
unless ::ActiveRecord::Base.connection_pool.respond_to?(:stat)
|
12
|
-
|
14
|
+
client.logger.error("ActiveRecord connection pool stats not supported in your rails version")
|
13
15
|
return
|
14
16
|
end
|
15
17
|
|
@@ -18,8 +20,6 @@ module PrometheusExporter::Instrumentation
|
|
18
20
|
|
19
21
|
active_record_collector = new(custom_labels, config_labels)
|
20
22
|
|
21
|
-
client ||= PrometheusExporter::Client.default
|
22
|
-
|
23
23
|
stop if @thread
|
24
24
|
|
25
25
|
@thread = Thread.new do
|
@@ -28,7 +28,7 @@ module PrometheusExporter::Instrumentation
|
|
28
28
|
metrics = active_record_collector.collect
|
29
29
|
metrics.each { |metric| client.send_json metric }
|
30
30
|
rescue => e
|
31
|
-
|
31
|
+
client.logger.error("Prometheus Exporter Failed To Collect Process Stats #{e}")
|
32
32
|
ensure
|
33
33
|
sleep frequency
|
34
34
|
end
|
@@ -67,21 +67,28 @@ module PrometheusExporter::Instrumentation
|
|
67
67
|
ObjectSpace.each_object(::ActiveRecord::ConnectionAdapters::ConnectionPool) do |pool|
|
68
68
|
next if pool.connections.nil?
|
69
69
|
|
70
|
-
labels_from_config = pool.spec.config
|
71
|
-
.select { |k, v| @config_labels.include? k }
|
72
|
-
.map { |k, v| [k.to_s.prepend("dbconfig_"), v] }
|
73
|
-
|
74
|
-
labels = @metric_labels.merge(pool_name: pool.spec.name).merge(Hash[labels_from_config])
|
75
|
-
|
76
70
|
metric = {
|
77
71
|
pid: pid,
|
78
72
|
type: "active_record",
|
79
73
|
hostname: ::PrometheusExporter.hostname,
|
80
|
-
metric_labels: labels
|
74
|
+
metric_labels: labels(pool)
|
81
75
|
}
|
82
76
|
metric.merge!(pool.stat)
|
83
77
|
metrics << metric
|
84
78
|
end
|
85
79
|
end
|
80
|
+
|
81
|
+
private
|
82
|
+
|
83
|
+
def labels(pool)
|
84
|
+
if ::ActiveRecord.version < Gem::Version.new("6.1.0.rc1")
|
85
|
+
@metric_labels.merge(pool_name: pool.spec.name).merge(pool.spec.config
|
86
|
+
.select { |k, v| @config_labels.include? k }
|
87
|
+
.map { |k, v| [k.to_s.dup.prepend("dbconfig_"), v] }.to_h)
|
88
|
+
else
|
89
|
+
@metric_labels.merge(pool_name: pool.db_config.name).merge(
|
90
|
+
@config_labels.each_with_object({}) { |l, acc| acc["dbconfig_#{l}"] = pool.db_config.public_send(l) })
|
91
|
+
end
|
92
|
+
end
|
86
93
|
end
|
87
94
|
end
|
@@ -13,8 +13,8 @@ module PrometheusExporter::Instrumentation
|
|
13
13
|
callbacks do |lifecycle|
|
14
14
|
lifecycle.around(:invoke_job) do |job, *args, &block|
|
15
15
|
max_attempts = Delayed::Worker.max_attempts
|
16
|
-
enqueued_count = Delayed::Job.count
|
17
|
-
pending_count = Delayed::Job.where(attempts: 0, locked_at: nil).count
|
16
|
+
enqueued_count = Delayed::Job.where(queue: job.queue).count
|
17
|
+
pending_count = Delayed::Job.where(attempts: 0, locked_at: nil, queue: job.queue).count
|
18
18
|
instrumenter.call(job, max_attempts, enqueued_count, pending_count, *args, &block)
|
19
19
|
end
|
20
20
|
end
|
@@ -41,6 +41,7 @@ module PrometheusExporter::Instrumentation
|
|
41
41
|
@client.send_json(
|
42
42
|
type: "delayed_job",
|
43
43
|
name: job.handler.to_s.match(JOB_CLASS_REGEXP).to_a[1].to_s,
|
44
|
+
queue_name: job.queue,
|
44
45
|
success: success,
|
45
46
|
duration: duration,
|
46
47
|
attempts: attempts,
|
@@ -5,6 +5,7 @@ module PrometheusExporter::Instrumentation; end
|
|
5
5
|
|
6
6
|
class PrometheusExporter::Instrumentation::MethodProfiler
|
7
7
|
def self.patch(klass, methods, name)
|
8
|
+
patch_source_line = __LINE__ + 3
|
8
9
|
patches = methods.map do |method_name|
|
9
10
|
<<~RUBY
|
10
11
|
unless defined?(#{method_name}__mp_unpatched)
|
@@ -26,7 +27,7 @@ class PrometheusExporter::Instrumentation::MethodProfiler
|
|
26
27
|
RUBY
|
27
28
|
end.join("\n")
|
28
29
|
|
29
|
-
klass.class_eval patches
|
30
|
+
klass.class_eval patches, __FILE__, patch_source_line
|
30
31
|
end
|
31
32
|
|
32
33
|
def self.transfer
|
@@ -27,7 +27,7 @@ module PrometheusExporter::Instrumentation
|
|
27
27
|
metric = process_collector.collect
|
28
28
|
client.send_json metric
|
29
29
|
rescue => e
|
30
|
-
|
30
|
+
client.logger.error("Prometheus Exporter Failed To Collect Process Stats #{e}")
|
31
31
|
ensure
|
32
32
|
sleep frequency
|
33
33
|
end
|
@@ -5,8 +5,8 @@ require "json"
|
|
5
5
|
# collects stats from puma
|
6
6
|
module PrometheusExporter::Instrumentation
|
7
7
|
class Puma
|
8
|
-
def self.start(client: nil, frequency: 30)
|
9
|
-
puma_collector = new
|
8
|
+
def self.start(client: nil, frequency: 30, labels: {})
|
9
|
+
puma_collector = new(labels)
|
10
10
|
client ||= PrometheusExporter::Client.default
|
11
11
|
Thread.new do
|
12
12
|
while true
|
@@ -14,7 +14,7 @@ module PrometheusExporter::Instrumentation
|
|
14
14
|
metric = puma_collector.collect
|
15
15
|
client.send_json metric
|
16
16
|
rescue => e
|
17
|
-
|
17
|
+
client.logger.error("Prometheus Exporter Failed To Collect Puma Stats #{e}")
|
18
18
|
ensure
|
19
19
|
sleep frequency
|
20
20
|
end
|
@@ -22,13 +22,25 @@ module PrometheusExporter::Instrumentation
|
|
22
22
|
end
|
23
23
|
end
|
24
24
|
|
25
|
+
def initialize(metric_labels = {})
|
26
|
+
@metric_labels = metric_labels
|
27
|
+
end
|
28
|
+
|
25
29
|
def collect
|
26
|
-
metric = {
|
27
|
-
|
30
|
+
metric = {
|
31
|
+
pid: pid,
|
32
|
+
type: "puma",
|
33
|
+
hostname: ::PrometheusExporter.hostname,
|
34
|
+
metric_labels: @metric_labels
|
35
|
+
}
|
28
36
|
collect_puma_stats(metric)
|
29
37
|
metric
|
30
38
|
end
|
31
39
|
|
40
|
+
def pid
|
41
|
+
@pid = ::Process.pid
|
42
|
+
end
|
43
|
+
|
32
44
|
def collect_puma_stats(metric)
|
33
45
|
stats = JSON.parse(::Puma.stats)
|
34
46
|
|
@@ -0,0 +1,40 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
# collects stats from resque
|
4
|
+
module PrometheusExporter::Instrumentation
|
5
|
+
class Resque
|
6
|
+
def self.start(client: nil, frequency: 30)
|
7
|
+
resque_collector = new
|
8
|
+
client ||= PrometheusExporter::Client.default
|
9
|
+
Thread.new do
|
10
|
+
while true
|
11
|
+
begin
|
12
|
+
client.send_json(resque_collector.collect)
|
13
|
+
rescue => e
|
14
|
+
client.logger.error("Prometheus Exporter Failed To Collect Resque Stats #{e}")
|
15
|
+
ensure
|
16
|
+
sleep frequency
|
17
|
+
end
|
18
|
+
end
|
19
|
+
end
|
20
|
+
end
|
21
|
+
|
22
|
+
def collect
|
23
|
+
metric = {}
|
24
|
+
metric[:type] = "resque"
|
25
|
+
collect_resque_stats(metric)
|
26
|
+
metric
|
27
|
+
end
|
28
|
+
|
29
|
+
def collect_resque_stats(metric)
|
30
|
+
info = ::Resque.info
|
31
|
+
|
32
|
+
metric[:processed_jobs_total] = info[:processed]
|
33
|
+
metric[:failed_jobs_total] = info[:failed]
|
34
|
+
metric[:pending_jobs_total] = info[:pending]
|
35
|
+
metric[:queues_total] = info[:queues]
|
36
|
+
metric[:worker_total] = info[:workers]
|
37
|
+
metric[:working_total] = info[:working]
|
38
|
+
end
|
39
|
+
end
|
40
|
+
end
|
@@ -1,6 +1,15 @@
|
|
1
1
|
# frozen_string_literal: true
|
2
2
|
|
3
|
+
require 'yaml'
|
4
|
+
|
3
5
|
module PrometheusExporter::Instrumentation
|
6
|
+
JOB_WRAPPER_CLASS_NAME = 'ActiveJob::QueueAdapters::SidekiqAdapter::JobWrapper'
|
7
|
+
DELAYED_CLASS_NAMES = [
|
8
|
+
'Sidekiq::Extensions::DelayedClass',
|
9
|
+
'Sidekiq::Extensions::DelayedModel',
|
10
|
+
'Sidekiq::Extensions::DelayedMailer',
|
11
|
+
]
|
12
|
+
|
4
13
|
class Sidekiq
|
5
14
|
def self.death_handler
|
6
15
|
-> (job, ex) do
|
@@ -32,15 +41,47 @@ module PrometheusExporter::Instrumentation
|
|
32
41
|
raise e
|
33
42
|
ensure
|
34
43
|
duration = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC) - start
|
35
|
-
class_name = worker.class.to_s == 'ActiveJob::QueueAdapters::SidekiqAdapter::JobWrapper' ?
|
36
|
-
msg['wrapped'] : worker.class.to_s
|
37
44
|
@client.send_json(
|
38
45
|
type: "sidekiq",
|
39
|
-
name:
|
46
|
+
name: get_name(worker, msg),
|
47
|
+
queue: queue,
|
40
48
|
success: success,
|
41
49
|
shutdown: shutdown,
|
42
50
|
duration: duration
|
43
51
|
)
|
44
52
|
end
|
53
|
+
|
54
|
+
private
|
55
|
+
|
56
|
+
def get_name(worker, msg)
|
57
|
+
class_name = worker.class.to_s
|
58
|
+
if class_name == JOB_WRAPPER_CLASS_NAME
|
59
|
+
get_job_wrapper_name(msg)
|
60
|
+
elsif DELAYED_CLASS_NAMES.include?(class_name)
|
61
|
+
get_delayed_name(msg, class_name)
|
62
|
+
else
|
63
|
+
class_name
|
64
|
+
end
|
65
|
+
end
|
66
|
+
|
67
|
+
def get_job_wrapper_name(msg)
|
68
|
+
msg['wrapped']
|
69
|
+
end
|
70
|
+
|
71
|
+
def get_delayed_name(msg, class_name)
|
72
|
+
# fallback to class_name since we're relying on the internal implementation
|
73
|
+
# of the delayed extensions
|
74
|
+
# https://github.com/mperham/sidekiq/blob/master/lib/sidekiq/extensions/class_methods.rb
|
75
|
+
begin
|
76
|
+
(target, method_name, _args) = YAML.load(msg['args'].first)
|
77
|
+
if target.class == Class
|
78
|
+
"#{target.name}##{method_name}"
|
79
|
+
else
|
80
|
+
"#{target.class.name}##{method_name}"
|
81
|
+
end
|
82
|
+
rescue
|
83
|
+
class_name
|
84
|
+
end
|
85
|
+
end
|
45
86
|
end
|
46
87
|
end
|
@@ -11,7 +11,7 @@ module PrometheusExporter::Instrumentation
|
|
11
11
|
begin
|
12
12
|
client.send_json(sidekiq_queue_collector.collect)
|
13
13
|
rescue StandardError => e
|
14
|
-
|
14
|
+
client.logger.error("Prometheus Exporter Failed To Collect Sidekiq Queue metrics #{e}")
|
15
15
|
ensure
|
16
16
|
sleep frequency
|
17
17
|
end
|
@@ -27,13 +27,24 @@ module PrometheusExporter::Instrumentation
|
|
27
27
|
end
|
28
28
|
|
29
29
|
def collect_queue_stats
|
30
|
+
hostname = Socket.gethostname
|
31
|
+
pid = ::Process.pid
|
32
|
+
ps = ::Sidekiq::ProcessSet.new
|
33
|
+
|
34
|
+
process = ps.find do |sp|
|
35
|
+
sp['hostname'] == hostname && sp['pid'] == pid
|
36
|
+
end
|
37
|
+
|
38
|
+
queues = process.nil? ? [] : process['queues']
|
39
|
+
|
30
40
|
::Sidekiq::Queue.all.map do |queue|
|
41
|
+
next unless queues.include? queue.name
|
31
42
|
{
|
32
43
|
backlog_total: queue.size,
|
33
44
|
latency_seconds: queue.latency.to_i,
|
34
45
|
labels: { queue: queue.name }
|
35
46
|
}
|
36
|
-
end
|
47
|
+
end.compact
|
37
48
|
end
|
38
49
|
end
|
39
50
|
end
|