karafka 1.4.0 → 2.0.10
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- checksums.yaml.gz.sig +0 -0
- data/.github/workflows/ci.yml +89 -18
- data/.ruby-version +1 -1
- data/CHANGELOG.md +365 -1
- data/CONTRIBUTING.md +10 -19
- data/Gemfile +6 -0
- data/Gemfile.lock +56 -112
- data/LICENSE +17 -0
- data/LICENSE-COMM +89 -0
- data/LICENSE-LGPL +165 -0
- data/README.md +61 -68
- data/bin/benchmarks +85 -0
- data/bin/create_token +22 -0
- data/bin/integrations +272 -0
- data/bin/karafka +10 -0
- data/bin/scenario +29 -0
- data/bin/stress_many +13 -0
- data/bin/stress_one +13 -0
- data/certs/cert_chain.pem +26 -0
- data/certs/karafka-pro.pem +11 -0
- data/config/errors.yml +59 -38
- data/docker-compose.yml +10 -3
- data/karafka.gemspec +18 -21
- data/lib/active_job/karafka.rb +21 -0
- data/lib/active_job/queue_adapters/karafka_adapter.rb +26 -0
- data/lib/karafka/active_job/consumer.rb +26 -0
- data/lib/karafka/active_job/dispatcher.rb +38 -0
- data/lib/karafka/active_job/job_extensions.rb +34 -0
- data/lib/karafka/active_job/job_options_contract.rb +21 -0
- data/lib/karafka/active_job/routing/extensions.rb +33 -0
- data/lib/karafka/admin.rb +63 -0
- data/lib/karafka/app.rb +15 -20
- data/lib/karafka/base_consumer.rb +197 -31
- data/lib/karafka/cli/info.rb +44 -10
- data/lib/karafka/cli/install.rb +22 -12
- data/lib/karafka/cli/server.rb +17 -42
- data/lib/karafka/cli.rb +4 -3
- data/lib/karafka/connection/client.rb +379 -89
- data/lib/karafka/connection/listener.rb +250 -38
- data/lib/karafka/connection/listeners_batch.rb +24 -0
- data/lib/karafka/connection/messages_buffer.rb +84 -0
- data/lib/karafka/connection/pauses_manager.rb +46 -0
- data/lib/karafka/connection/raw_messages_buffer.rb +101 -0
- data/lib/karafka/connection/rebalance_manager.rb +78 -0
- data/lib/karafka/contracts/base.rb +17 -0
- data/lib/karafka/contracts/config.rb +88 -11
- data/lib/karafka/contracts/consumer_group.rb +21 -184
- data/lib/karafka/contracts/consumer_group_topic.rb +35 -11
- data/lib/karafka/contracts/server_cli_options.rb +19 -18
- data/lib/karafka/contracts.rb +1 -1
- data/lib/karafka/env.rb +46 -0
- data/lib/karafka/errors.rb +21 -21
- data/lib/karafka/helpers/async.rb +33 -0
- data/lib/karafka/helpers/colorize.rb +20 -0
- data/lib/karafka/helpers/multi_delegator.rb +2 -2
- data/lib/karafka/instrumentation/callbacks/error.rb +40 -0
- data/lib/karafka/instrumentation/callbacks/statistics.rb +41 -0
- data/lib/karafka/instrumentation/logger.rb +6 -10
- data/lib/karafka/instrumentation/logger_listener.rb +174 -0
- data/lib/karafka/instrumentation/monitor.rb +13 -61
- data/lib/karafka/instrumentation/notifications.rb +53 -0
- data/lib/karafka/instrumentation/proctitle_listener.rb +3 -3
- data/lib/karafka/instrumentation/vendors/datadog/dashboard.json +1 -0
- data/lib/karafka/instrumentation/vendors/datadog/listener.rb +232 -0
- data/lib/karafka/instrumentation.rb +21 -0
- data/lib/karafka/licenser.rb +75 -0
- data/lib/karafka/messages/batch_metadata.rb +45 -0
- data/lib/karafka/messages/builders/batch_metadata.rb +39 -0
- data/lib/karafka/messages/builders/message.rb +39 -0
- data/lib/karafka/messages/builders/messages.rb +34 -0
- data/lib/karafka/{params/params.rb → messages/message.rb} +7 -12
- data/lib/karafka/messages/messages.rb +64 -0
- data/lib/karafka/{params → messages}/metadata.rb +4 -6
- data/lib/karafka/messages/seek.rb +9 -0
- data/lib/karafka/patches/rdkafka/consumer.rb +22 -0
- data/lib/karafka/pro/active_job/consumer.rb +46 -0
- data/lib/karafka/pro/active_job/dispatcher.rb +61 -0
- data/lib/karafka/pro/active_job/job_options_contract.rb +32 -0
- data/lib/karafka/pro/base_consumer.rb +107 -0
- data/lib/karafka/pro/contracts/base.rb +21 -0
- data/lib/karafka/pro/contracts/consumer_group.rb +34 -0
- data/lib/karafka/pro/contracts/consumer_group_topic.rb +69 -0
- data/lib/karafka/pro/loader.rb +76 -0
- data/lib/karafka/pro/performance_tracker.rb +80 -0
- data/lib/karafka/pro/processing/coordinator.rb +85 -0
- data/lib/karafka/pro/processing/jobs/consume_non_blocking.rb +38 -0
- data/lib/karafka/pro/processing/jobs_builder.rb +32 -0
- data/lib/karafka/pro/processing/partitioner.rb +58 -0
- data/lib/karafka/pro/processing/scheduler.rb +56 -0
- data/lib/karafka/pro/routing/builder_extensions.rb +30 -0
- data/lib/karafka/pro/routing/topic_extensions.rb +74 -0
- data/lib/karafka/pro.rb +13 -0
- data/lib/karafka/process.rb +1 -0
- data/lib/karafka/processing/coordinator.rb +103 -0
- data/lib/karafka/processing/coordinators_buffer.rb +54 -0
- data/lib/karafka/processing/executor.rb +126 -0
- data/lib/karafka/processing/executors_buffer.rb +88 -0
- data/lib/karafka/processing/jobs/base.rb +55 -0
- data/lib/karafka/processing/jobs/consume.rb +47 -0
- data/lib/karafka/processing/jobs/revoked.rb +22 -0
- data/lib/karafka/processing/jobs/shutdown.rb +23 -0
- data/lib/karafka/processing/jobs_builder.rb +29 -0
- data/lib/karafka/processing/jobs_queue.rb +144 -0
- data/lib/karafka/processing/partitioner.rb +22 -0
- data/lib/karafka/processing/result.rb +37 -0
- data/lib/karafka/processing/scheduler.rb +22 -0
- data/lib/karafka/processing/worker.rb +91 -0
- data/lib/karafka/processing/workers_batch.rb +27 -0
- data/lib/karafka/railtie.rb +127 -0
- data/lib/karafka/routing/builder.rb +26 -23
- data/lib/karafka/routing/consumer_group.rb +37 -17
- data/lib/karafka/routing/consumer_mapper.rb +1 -2
- data/lib/karafka/routing/proxy.rb +9 -16
- data/lib/karafka/routing/router.rb +1 -1
- data/lib/karafka/routing/subscription_group.rb +53 -0
- data/lib/karafka/routing/subscription_groups_builder.rb +54 -0
- data/lib/karafka/routing/topic.rb +65 -24
- data/lib/karafka/routing/topics.rb +38 -0
- data/lib/karafka/runner.rb +51 -0
- data/lib/karafka/serialization/json/deserializer.rb +6 -15
- data/lib/karafka/server.rb +67 -26
- data/lib/karafka/setup/config.rb +153 -175
- data/lib/karafka/status.rb +14 -5
- data/lib/karafka/templates/example_consumer.rb.erb +16 -0
- data/lib/karafka/templates/karafka.rb.erb +17 -55
- data/lib/karafka/time_trackers/base.rb +19 -0
- data/lib/karafka/time_trackers/pause.rb +92 -0
- data/lib/karafka/time_trackers/poll.rb +65 -0
- data/lib/karafka/version.rb +1 -1
- data/lib/karafka.rb +46 -16
- data.tar.gz.sig +0 -0
- metadata +145 -171
- metadata.gz.sig +0 -0
- data/.github/FUNDING.yml +0 -3
- data/MIT-LICENCE +0 -18
- data/certs/mensfeld.pem +0 -25
- data/lib/karafka/attributes_map.rb +0 -62
- data/lib/karafka/backends/inline.rb +0 -16
- data/lib/karafka/base_responder.rb +0 -226
- data/lib/karafka/cli/flow.rb +0 -48
- data/lib/karafka/code_reloader.rb +0 -67
- data/lib/karafka/connection/api_adapter.rb +0 -161
- data/lib/karafka/connection/batch_delegator.rb +0 -55
- data/lib/karafka/connection/builder.rb +0 -18
- data/lib/karafka/connection/message_delegator.rb +0 -36
- data/lib/karafka/consumers/batch_metadata.rb +0 -10
- data/lib/karafka/consumers/callbacks.rb +0 -71
- data/lib/karafka/consumers/includer.rb +0 -64
- data/lib/karafka/consumers/responders.rb +0 -24
- data/lib/karafka/consumers/single_params.rb +0 -15
- data/lib/karafka/contracts/responder_usage.rb +0 -54
- data/lib/karafka/fetcher.rb +0 -42
- data/lib/karafka/helpers/class_matcher.rb +0 -88
- data/lib/karafka/helpers/config_retriever.rb +0 -46
- data/lib/karafka/helpers/inflector.rb +0 -26
- data/lib/karafka/instrumentation/stdout_listener.rb +0 -140
- data/lib/karafka/params/batch_metadata.rb +0 -26
- data/lib/karafka/params/builders/batch_metadata.rb +0 -30
- data/lib/karafka/params/builders/params.rb +0 -38
- data/lib/karafka/params/builders/params_batch.rb +0 -25
- data/lib/karafka/params/params_batch.rb +0 -60
- data/lib/karafka/patches/ruby_kafka.rb +0 -47
- data/lib/karafka/persistence/client.rb +0 -29
- data/lib/karafka/persistence/consumers.rb +0 -45
- data/lib/karafka/persistence/topics.rb +0 -48
- data/lib/karafka/responders/builder.rb +0 -36
- data/lib/karafka/responders/topic.rb +0 -55
- data/lib/karafka/routing/topic_mapper.rb +0 -53
- data/lib/karafka/serialization/json/serializer.rb +0 -31
- data/lib/karafka/setup/configurators/water_drop.rb +0 -36
- data/lib/karafka/templates/application_responder.rb.erb +0 -11
@@ -0,0 +1,47 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Karafka
|
4
|
+
module Processing
|
5
|
+
module Jobs
|
6
|
+
# The main job type. It runs the executor that triggers given topic partition messages
|
7
|
+
# processing in an underlying consumer instance.
|
8
|
+
class Consume < Base
|
9
|
+
# @return [Array<Rdkafka::Consumer::Message>] array with messages
|
10
|
+
attr_reader :messages
|
11
|
+
|
12
|
+
# @param executor [Karafka::Processing::Executor] executor that is suppose to run a given
|
13
|
+
# job
|
14
|
+
# @param messages [Karafka::Messages::Messages] karafka messages batch
|
15
|
+
# @param coordinator [Karafka::Processing::Coordinator] processing coordinator
|
16
|
+
# @return [Consume]
|
17
|
+
def initialize(executor, messages, coordinator)
|
18
|
+
@executor = executor
|
19
|
+
@messages = messages
|
20
|
+
@coordinator = coordinator
|
21
|
+
super()
|
22
|
+
end
|
23
|
+
|
24
|
+
# Runs all the preparation code on the executor that needs to happen before the job is
|
25
|
+
# enqueued.
|
26
|
+
def before_enqueue
|
27
|
+
executor.before_enqueue(@messages, @coordinator)
|
28
|
+
end
|
29
|
+
|
30
|
+
# Runs the before consumption preparations on the executor
|
31
|
+
def before_call
|
32
|
+
executor.before_consume
|
33
|
+
end
|
34
|
+
|
35
|
+
# Runs the given executor
|
36
|
+
def call
|
37
|
+
executor.consume
|
38
|
+
end
|
39
|
+
|
40
|
+
# Runs any error handling and other post-consumption stuff on the executor
|
41
|
+
def after_call
|
42
|
+
executor.after_consume
|
43
|
+
end
|
44
|
+
end
|
45
|
+
end
|
46
|
+
end
|
47
|
+
end
|
@@ -0,0 +1,22 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Karafka
|
4
|
+
module Processing
|
5
|
+
module Jobs
|
6
|
+
# Job that runs the revoked operation when we loose a partition on a consumer that lost it.
|
7
|
+
class Revoked < Base
|
8
|
+
# @param executor [Karafka::Processing::Executor] executor that is suppose to run the job
|
9
|
+
# @return [Revoked]
|
10
|
+
def initialize(executor)
|
11
|
+
@executor = executor
|
12
|
+
super()
|
13
|
+
end
|
14
|
+
|
15
|
+
# Runs the revoking job via an executor.
|
16
|
+
def call
|
17
|
+
executor.revoked
|
18
|
+
end
|
19
|
+
end
|
20
|
+
end
|
21
|
+
end
|
22
|
+
end
|
@@ -0,0 +1,23 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Karafka
|
4
|
+
module Processing
|
5
|
+
module Jobs
|
6
|
+
# Job that runs on each active consumer upon process shutdown (one job per consumer).
|
7
|
+
class Shutdown < Base
|
8
|
+
# @param executor [Karafka::Processing::Executor] executor that is suppose to run a given
|
9
|
+
# job on an active consumer
|
10
|
+
# @return [Shutdown]
|
11
|
+
def initialize(executor)
|
12
|
+
@executor = executor
|
13
|
+
super()
|
14
|
+
end
|
15
|
+
|
16
|
+
# Runs the shutdown job via an executor.
|
17
|
+
def call
|
18
|
+
executor.shutdown
|
19
|
+
end
|
20
|
+
end
|
21
|
+
end
|
22
|
+
end
|
23
|
+
end
|
@@ -0,0 +1,29 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Karafka
|
4
|
+
module Processing
|
5
|
+
# Class responsible for deciding what type of job should we build to run a given command and
|
6
|
+
# for building a proper job for it.
|
7
|
+
class JobsBuilder
|
8
|
+
# @param executor [Karafka::Processing::Executor]
|
9
|
+
# @param messages [Karafka::Messages::Messages] messages batch to be consumed
|
10
|
+
# @param coordinator [Karafka::Processing::Coordinator]
|
11
|
+
# @return [Karafka::Processing::Jobs::Consume] consumption job
|
12
|
+
def consume(executor, messages, coordinator)
|
13
|
+
Jobs::Consume.new(executor, messages, coordinator)
|
14
|
+
end
|
15
|
+
|
16
|
+
# @param executor [Karafka::Processing::Executor]
|
17
|
+
# @return [Karafka::Processing::Jobs::Revoked] revocation job
|
18
|
+
def revoked(executor)
|
19
|
+
Jobs::Revoked.new(executor)
|
20
|
+
end
|
21
|
+
|
22
|
+
# @param executor [Karafka::Processing::Executor]
|
23
|
+
# @return [Karafka::Processing::Jobs::Shutdown] shutdown job
|
24
|
+
def shutdown(executor)
|
25
|
+
Jobs::Shutdown.new(executor)
|
26
|
+
end
|
27
|
+
end
|
28
|
+
end
|
29
|
+
end
|
@@ -0,0 +1,144 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Karafka
|
4
|
+
module Processing
|
5
|
+
# This is the key work component for Karafka jobs distribution. It provides API for running
|
6
|
+
# jobs in parallel while operating within more than one subscription group.
|
7
|
+
#
|
8
|
+
# We need to take into consideration fact, that more than one subscription group can operate
|
9
|
+
# on this queue, that's why internally we keep track of processing per group.
|
10
|
+
#
|
11
|
+
# We work with the assumption, that partitions data is evenly distributed.
|
12
|
+
class JobsQueue
|
13
|
+
# @return [Karafka::Processing::JobsQueue]
|
14
|
+
def initialize
|
15
|
+
@queue = Queue.new
|
16
|
+
# Those queues will act as semaphores internally. Since we need an indicator for waiting
|
17
|
+
# we could use Thread.pass but this is expensive. Instead we can just lock until any
|
18
|
+
# of the workers finishes their work and we can re-check. This means that in the worse
|
19
|
+
# scenario, we will context switch 10 times per poll instead of getting this thread
|
20
|
+
# scheduled by Ruby hundreds of thousands of times per group.
|
21
|
+
# We cannot use a single semaphore as it could potentially block in listeners that should
|
22
|
+
# process with their data and also could unlock when a given group needs to remain locked
|
23
|
+
@semaphores = Hash.new { |h, k| h[k] = Queue.new }
|
24
|
+
@in_processing = Hash.new { |h, k| h[k] = [] }
|
25
|
+
@mutex = Mutex.new
|
26
|
+
end
|
27
|
+
|
28
|
+
# Returns number of jobs that are either enqueued or in processing (but not finished)
|
29
|
+
# @return [Integer] number of elements in the queue
|
30
|
+
# @note Using `#pop` won't decrease this number as only marking job as completed does this
|
31
|
+
def size
|
32
|
+
@in_processing.values.map(&:size).sum
|
33
|
+
end
|
34
|
+
|
35
|
+
# Adds the job to the internal main queue, scheduling it for execution in a worker and marks
|
36
|
+
# this job as in processing pipeline.
|
37
|
+
#
|
38
|
+
# @param job [Jobs::Base] job that we want to run
|
39
|
+
def <<(job)
|
40
|
+
# We do not push the job if the queue is closed as it means that it would anyhow not be
|
41
|
+
# executed
|
42
|
+
return if @queue.closed?
|
43
|
+
|
44
|
+
@mutex.synchronize do
|
45
|
+
group = @in_processing[job.group_id]
|
46
|
+
|
47
|
+
raise(Errors::JobsQueueSynchronizationError, job.group_id) if group.include?(job)
|
48
|
+
|
49
|
+
group << job
|
50
|
+
end
|
51
|
+
|
52
|
+
@queue << job
|
53
|
+
end
|
54
|
+
|
55
|
+
# @return [Jobs::Base, nil] waits for a job from the main queue and returns it once available
|
56
|
+
# or returns nil if the queue has been stopped and there won't be anything more to process
|
57
|
+
# ever.
|
58
|
+
# @note This command is blocking and will wait until any job is available on the main queue
|
59
|
+
def pop
|
60
|
+
@queue.pop
|
61
|
+
end
|
62
|
+
|
63
|
+
# Causes the wait lock to re-check the lock conditions and potential unlock.
|
64
|
+
# @param group_id [String] id of the group we want to unlock for one tick
|
65
|
+
# @note This does not release the wait lock. It just causes a conditions recheck
|
66
|
+
def tick(group_id)
|
67
|
+
@semaphores[group_id] << true
|
68
|
+
end
|
69
|
+
|
70
|
+
# Marks a given job from a given group as completed. When there are no more jobs from a given
|
71
|
+
# group to be executed, we won't wait.
|
72
|
+
#
|
73
|
+
# @param [Jobs::Base] job that was completed
|
74
|
+
def complete(job)
|
75
|
+
@mutex.synchronize do
|
76
|
+
@in_processing[job.group_id].delete(job)
|
77
|
+
tick(job.group_id)
|
78
|
+
end
|
79
|
+
end
|
80
|
+
|
81
|
+
# Clears the processing states for a provided group. Useful when a recovery happens and we
|
82
|
+
# need to clean up state but only for a given subscription group.
|
83
|
+
#
|
84
|
+
# @param group_id [String]
|
85
|
+
def clear(group_id)
|
86
|
+
@mutex.synchronize do
|
87
|
+
@in_processing[group_id].clear
|
88
|
+
# We unlock it just in case it was blocked when clearing started
|
89
|
+
tick(group_id)
|
90
|
+
end
|
91
|
+
end
|
92
|
+
|
93
|
+
# Stops the whole processing queue.
|
94
|
+
def close
|
95
|
+
@mutex.synchronize do
|
96
|
+
return if @queue.closed?
|
97
|
+
|
98
|
+
@queue.close
|
99
|
+
@semaphores.values.each(&:close)
|
100
|
+
end
|
101
|
+
end
|
102
|
+
|
103
|
+
# @param group_id [String]
|
104
|
+
#
|
105
|
+
# @return [Boolean] tell us if we have anything in the processing (or for processing) from
|
106
|
+
# a given group.
|
107
|
+
def empty?(group_id)
|
108
|
+
@in_processing[group_id].empty?
|
109
|
+
end
|
110
|
+
|
111
|
+
# Blocks when there are things in the queue in a given group and waits until all the blocking
|
112
|
+
# jobs from a given group are completed
|
113
|
+
#
|
114
|
+
# @param group_id [String] id of the group in which jobs we're interested.
|
115
|
+
# @note This method is blocking.
|
116
|
+
def wait(group_id)
|
117
|
+
# Go doing other things while we cannot process and wait for anyone to finish their work
|
118
|
+
# and re-check the wait status
|
119
|
+
@semaphores[group_id].pop while wait?(group_id)
|
120
|
+
end
|
121
|
+
|
122
|
+
# - `processing` - number of jobs that are currently being processed (active work)
|
123
|
+
# - `enqueued` - number of jobs in the queue that are waiting to be picked up by a worker
|
124
|
+
#
|
125
|
+
# @return [Hash] hash with basic usage statistics of this queue.
|
126
|
+
def statistics
|
127
|
+
{
|
128
|
+
processing: size - @queue.size,
|
129
|
+
enqueued: @queue.size
|
130
|
+
}.freeze
|
131
|
+
end
|
132
|
+
|
133
|
+
private
|
134
|
+
|
135
|
+
# @param group_id [String] id of the group in which jobs we're interested.
|
136
|
+
# @return [Boolean] should we keep waiting or not
|
137
|
+
# @note We do not wait for non-blocking jobs. Their flow should allow for `poll` running
|
138
|
+
# as they may exceed `max.poll.interval`
|
139
|
+
def wait?(group_id)
|
140
|
+
!@in_processing[group_id].all?(&:non_blocking?)
|
141
|
+
end
|
142
|
+
end
|
143
|
+
end
|
144
|
+
end
|
@@ -0,0 +1,22 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Karafka
|
4
|
+
module Processing
|
5
|
+
# Basic partitioner for work division
|
6
|
+
# It does not divide any work.
|
7
|
+
class Partitioner
|
8
|
+
# @param subscription_group [Karafka::Routing::SubscriptionGroup] subscription group
|
9
|
+
def initialize(subscription_group)
|
10
|
+
@subscription_group = subscription_group
|
11
|
+
end
|
12
|
+
|
13
|
+
# @param _topic [String] topic name
|
14
|
+
# @param messages [Array<Karafka::Messages::Message>] karafka messages
|
15
|
+
# @yieldparam [Integer] group id
|
16
|
+
# @yieldparam [Array<Karafka::Messages::Message>] karafka messages
|
17
|
+
def call(_topic, messages)
|
18
|
+
yield(0, messages)
|
19
|
+
end
|
20
|
+
end
|
21
|
+
end
|
22
|
+
end
|
@@ -0,0 +1,37 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Karafka
|
4
|
+
module Processing
|
5
|
+
# A simple object that allows us to keep track of processing state.
|
6
|
+
# It allows to indicate if given thing moved from success to a failure or the other way around
|
7
|
+
# Useful for tracking consumption state
|
8
|
+
class Result
|
9
|
+
attr_reader :cause
|
10
|
+
|
11
|
+
def initialize
|
12
|
+
@success = true
|
13
|
+
@cause = false
|
14
|
+
end
|
15
|
+
|
16
|
+
# @return [Boolean]
|
17
|
+
def success?
|
18
|
+
@success
|
19
|
+
end
|
20
|
+
|
21
|
+
# Marks state as successful
|
22
|
+
def success!
|
23
|
+
@success = true
|
24
|
+
# We set cause to false so the previous error that occurred does not leak when error is
|
25
|
+
# no longer present
|
26
|
+
@cause = false
|
27
|
+
end
|
28
|
+
|
29
|
+
# Marks state as failure
|
30
|
+
# @param cause [StandardError] error that occurred and caused failure
|
31
|
+
def failure!(cause)
|
32
|
+
@success = false
|
33
|
+
@cause = cause
|
34
|
+
end
|
35
|
+
end
|
36
|
+
end
|
37
|
+
end
|
@@ -0,0 +1,22 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Karafka
|
4
|
+
module Processing
|
5
|
+
# FIFO scheduler for messages coming from various topics and partitions
|
6
|
+
class Scheduler
|
7
|
+
# Schedules jobs in the fifo order
|
8
|
+
#
|
9
|
+
# @param queue [Karafka::Processing::JobsQueue] queue where we want to put the jobs
|
10
|
+
# @param jobs_array [Array<Karafka::Processing::Jobs::Base>] jobs we want to schedule
|
11
|
+
def schedule_consumption(queue, jobs_array)
|
12
|
+
jobs_array.each do |job|
|
13
|
+
queue << job
|
14
|
+
end
|
15
|
+
end
|
16
|
+
|
17
|
+
# Both revocation and shutdown jobs can also run in fifo by default
|
18
|
+
alias schedule_revocation schedule_consumption
|
19
|
+
alias schedule_shutdown schedule_consumption
|
20
|
+
end
|
21
|
+
end
|
22
|
+
end
|
@@ -0,0 +1,91 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Karafka
|
4
|
+
module Processing
|
5
|
+
# Workers are used to run jobs in separate threads.
|
6
|
+
# Workers are the main processing units of the Karafka framework.
|
7
|
+
#
|
8
|
+
# Each job runs in three stages:
|
9
|
+
# - prepare - here we can run any code that we would need to run blocking before we allow
|
10
|
+
# the job to run fully async (non blocking). This will always run in a blocking
|
11
|
+
# way and can be used to make sure all the resources and external dependencies
|
12
|
+
# are satisfied before going async.
|
13
|
+
#
|
14
|
+
# - call - actual processing logic that can run sync or async
|
15
|
+
#
|
16
|
+
# - teardown - it should include any code that we want to run after we executed the user
|
17
|
+
# code. This can be used to unlock certain resources or do other things that are
|
18
|
+
# not user code but need to run after user code base is executed.
|
19
|
+
class Worker
|
20
|
+
include Helpers::Async
|
21
|
+
|
22
|
+
# @return [String] id of this worker
|
23
|
+
attr_reader :id
|
24
|
+
|
25
|
+
# @param jobs_queue [JobsQueue]
|
26
|
+
# @return [Worker]
|
27
|
+
def initialize(jobs_queue)
|
28
|
+
@id = SecureRandom.uuid
|
29
|
+
@jobs_queue = jobs_queue
|
30
|
+
end
|
31
|
+
|
32
|
+
private
|
33
|
+
|
34
|
+
# Runs processing of jobs in a loop
|
35
|
+
# Stops when queue is closed.
|
36
|
+
def call
|
37
|
+
loop { break unless process }
|
38
|
+
end
|
39
|
+
|
40
|
+
# Fetches a single job, processes it and marks as completed.
|
41
|
+
#
|
42
|
+
# @note We do not have error handling here, as no errors should propagate this far. If they
|
43
|
+
# do, it is a critical error and should bubble up.
|
44
|
+
#
|
45
|
+
# @note Upon closing the jobs queue, worker will close it's thread
|
46
|
+
def process
|
47
|
+
job = @jobs_queue.pop
|
48
|
+
|
49
|
+
instrument_details = { caller: self, job: job, jobs_queue: @jobs_queue }
|
50
|
+
|
51
|
+
if job
|
52
|
+
Karafka.monitor.instrument('worker.process', instrument_details)
|
53
|
+
|
54
|
+
Karafka.monitor.instrument('worker.processed', instrument_details) do
|
55
|
+
job.before_call
|
56
|
+
|
57
|
+
# If a job is marked as non blocking, we can run a tick in the job queue and if there
|
58
|
+
# are no other blocking factors, the job queue will be unlocked.
|
59
|
+
# If this does not run, all the things will be blocking and job queue won't allow to
|
60
|
+
# pass it until done.
|
61
|
+
@jobs_queue.tick(job.group_id) if job.non_blocking?
|
62
|
+
|
63
|
+
job.call
|
64
|
+
|
65
|
+
job.after_call
|
66
|
+
|
67
|
+
true
|
68
|
+
end
|
69
|
+
else
|
70
|
+
false
|
71
|
+
end
|
72
|
+
# We signal critical exceptions, notify and do not allow worker to fail
|
73
|
+
# rubocop:disable Lint/RescueException
|
74
|
+
rescue Exception => e
|
75
|
+
# rubocop:enable Lint/RescueException
|
76
|
+
Karafka.monitor.instrument(
|
77
|
+
'error.occurred',
|
78
|
+
caller: self,
|
79
|
+
error: e,
|
80
|
+
type: 'worker.process.error'
|
81
|
+
)
|
82
|
+
ensure
|
83
|
+
# job can be nil when the queue is being closed
|
84
|
+
@jobs_queue.complete(job) if job
|
85
|
+
|
86
|
+
# Always publish info, that we completed all the work despite its result
|
87
|
+
Karafka.monitor.instrument('worker.completed', instrument_details)
|
88
|
+
end
|
89
|
+
end
|
90
|
+
end
|
91
|
+
end
|
@@ -0,0 +1,27 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Karafka
|
4
|
+
module Processing
|
5
|
+
# Abstraction layer around workers batch.
|
6
|
+
class WorkersBatch
|
7
|
+
include Enumerable
|
8
|
+
|
9
|
+
# @param jobs_queue [JobsQueue]
|
10
|
+
# @return [WorkersBatch]
|
11
|
+
def initialize(jobs_queue)
|
12
|
+
@batch = Array.new(App.config.concurrency) { Processing::Worker.new(jobs_queue) }
|
13
|
+
end
|
14
|
+
|
15
|
+
# Iterates over available workers and yields each worker
|
16
|
+
# @param block [Proc] block we want to run
|
17
|
+
def each(&block)
|
18
|
+
@batch.each(&block)
|
19
|
+
end
|
20
|
+
|
21
|
+
# @return [Integer] number of workers in the batch
|
22
|
+
def size
|
23
|
+
@batch.size
|
24
|
+
end
|
25
|
+
end
|
26
|
+
end
|
27
|
+
end
|
@@ -0,0 +1,127 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
# This file contains Railtie for auto-configuration
|
4
|
+
|
5
|
+
rails = false
|
6
|
+
|
7
|
+
begin
|
8
|
+
require 'rails'
|
9
|
+
|
10
|
+
rails = true
|
11
|
+
rescue LoadError
|
12
|
+
# Without defining this in any way, Zeitwerk ain't happy so we do it that way
|
13
|
+
module Karafka
|
14
|
+
class Railtie
|
15
|
+
end
|
16
|
+
end
|
17
|
+
end
|
18
|
+
|
19
|
+
if rails
|
20
|
+
# Load Karafka
|
21
|
+
require 'karafka'
|
22
|
+
|
23
|
+
# Load ActiveJob adapter
|
24
|
+
require 'active_job/karafka'
|
25
|
+
|
26
|
+
# Setup env if configured (may be configured later by .net, etc)
|
27
|
+
ENV['KARAFKA_ENV'] ||= ENV['RAILS_ENV'] if ENV.key?('RAILS_ENV')
|
28
|
+
|
29
|
+
module Karafka
|
30
|
+
# Railtie for setting up Rails integration
|
31
|
+
class Railtie < Rails::Railtie
|
32
|
+
railtie_name :karafka
|
33
|
+
|
34
|
+
initializer 'karafka.active_job_integration' do
|
35
|
+
ActiveSupport.on_load(:active_job) do
|
36
|
+
# Extend ActiveJob with some Karafka specific ActiveJob magic
|
37
|
+
extend ::Karafka::ActiveJob::JobExtensions
|
38
|
+
end
|
39
|
+
end
|
40
|
+
|
41
|
+
# This lines will make Karafka print to stdout like puma or unicorn when we run karafka
|
42
|
+
# server + will support code reloading with each fetched loop. We do it only for karafka
|
43
|
+
# based commands as Rails processes and console will have it enabled already
|
44
|
+
initializer 'karafka.configure_rails_logger' do
|
45
|
+
# Make Karafka use Rails logger
|
46
|
+
::Karafka::App.config.logger = Rails.logger
|
47
|
+
|
48
|
+
next unless Rails.env.development?
|
49
|
+
next unless ENV.key?('KARAFKA_CLI')
|
50
|
+
|
51
|
+
logger = ActiveSupport::Logger.new($stdout)
|
52
|
+
# Inherit the logger level from Rails, otherwise would always run with the debug level
|
53
|
+
logger.level = Rails.logger.level
|
54
|
+
|
55
|
+
Rails.logger.extend(
|
56
|
+
ActiveSupport::Logger.broadcast(
|
57
|
+
logger
|
58
|
+
)
|
59
|
+
)
|
60
|
+
end
|
61
|
+
|
62
|
+
initializer 'karafka.configure_rails_auto_load_paths' do |app|
|
63
|
+
# Consumers should autoload by default in the Rails app so they are visible
|
64
|
+
app.config.autoload_paths += %w[app/consumers]
|
65
|
+
end
|
66
|
+
|
67
|
+
initializer 'karafka.configure_rails_code_reloader' do
|
68
|
+
# There are components that won't work with older Rails version, so we check it and
|
69
|
+
# provide a failover
|
70
|
+
rails6plus = Rails.gem_version >= Gem::Version.new('6.0.0')
|
71
|
+
|
72
|
+
next unless Rails.env.development?
|
73
|
+
next unless ENV.key?('KARAFKA_CLI')
|
74
|
+
next unless rails6plus
|
75
|
+
|
76
|
+
# We can have many listeners, but it does not matter in which we will reload the code
|
77
|
+
# as long as all the consumers will be re-created as Rails reload is thread-safe
|
78
|
+
::Karafka::App.monitor.subscribe('connection.listener.fetch_loop') do
|
79
|
+
# Reload code each time there is a change in the code
|
80
|
+
next unless Rails.application.reloaders.any?(&:updated?)
|
81
|
+
|
82
|
+
Rails.application.reloader.reload!
|
83
|
+
end
|
84
|
+
|
85
|
+
::Karafka::App.monitor.subscribe('worker.completed') do
|
86
|
+
# Skip in case someone is using Rails without ActiveRecord
|
87
|
+
next unless Object.const_defined?('ActiveRecord::Base')
|
88
|
+
|
89
|
+
# Always release the connection after processing is done. Otherwise thread may hang
|
90
|
+
# blocking the reload and further processing
|
91
|
+
# @see https://github.com/rails/rails/issues/44183
|
92
|
+
ActiveRecord::Base.connection_pool.release_connection
|
93
|
+
end
|
94
|
+
end
|
95
|
+
|
96
|
+
initializer 'karafka.require_karafka_boot_file' do |app|
|
97
|
+
rails6plus = Rails.gem_version >= Gem::Version.new('6.0.0')
|
98
|
+
|
99
|
+
# If the boot file location is set to "false", we should not raise an exception and we
|
100
|
+
# should just not load karafka stuff. Setting this explicitly to false indicates, that
|
101
|
+
# karafka is part of the supply chain but it is not a first class citizen of a given
|
102
|
+
# system (may be just a dependency of a dependency), thus railtie should not kick in to
|
103
|
+
# load the non-existing boot file
|
104
|
+
next if Karafka.boot_file.to_s == 'false'
|
105
|
+
|
106
|
+
karafka_boot_file = Rails.root.join(Karafka.boot_file.to_s).to_s
|
107
|
+
|
108
|
+
# Provide more comprehensive error for when no boot file
|
109
|
+
unless File.exist?(karafka_boot_file)
|
110
|
+
raise(Karafka::Errors::MissingBootFileError, karafka_boot_file)
|
111
|
+
end
|
112
|
+
|
113
|
+
if rails6plus
|
114
|
+
app.reloader.to_prepare do
|
115
|
+
# Load Karafka boot file, so it can be used in Rails server context
|
116
|
+
require karafka_boot_file
|
117
|
+
end
|
118
|
+
else
|
119
|
+
# Load Karafka main setup for older Rails versions
|
120
|
+
app.config.after_initialize do
|
121
|
+
require karafka_boot_file
|
122
|
+
end
|
123
|
+
end
|
124
|
+
end
|
125
|
+
end
|
126
|
+
end
|
127
|
+
end
|
@@ -10,13 +10,9 @@ module Karafka
|
|
10
10
|
# end
|
11
11
|
# end
|
12
12
|
class Builder < Concurrent::Array
|
13
|
-
# Consumer group consistency checking contract
|
14
|
-
CONTRACT = Karafka::Contracts::ConsumerGroup.new.freeze
|
15
|
-
|
16
|
-
private_constant :CONTRACT
|
17
|
-
|
18
13
|
def initialize
|
19
14
|
@draws = Concurrent::Array.new
|
15
|
+
super
|
20
16
|
end
|
21
17
|
|
22
18
|
# Used to draw routes for Karafka
|
@@ -37,11 +33,7 @@ module Karafka
|
|
37
33
|
instance_eval(&block)
|
38
34
|
|
39
35
|
each do |consumer_group|
|
40
|
-
|
41
|
-
validation_result = CONTRACT.call(hashed_group)
|
42
|
-
next if validation_result.success?
|
43
|
-
|
44
|
-
raise Errors::InvalidConfigurationError, validation_result.errors.to_h
|
36
|
+
Contracts::ConsumerGroup.new.validate!(consumer_group.to_h)
|
45
37
|
end
|
46
38
|
end
|
47
39
|
|
@@ -58,30 +50,41 @@ module Karafka
|
|
58
50
|
super
|
59
51
|
end
|
60
52
|
|
61
|
-
# Redraws all the routes for the in-process code reloading.
|
62
|
-
# @note This won't allow registration of new topics without process restart but will trigger
|
63
|
-
# cache invalidation so all the classes, etc are re-fetched after code reload
|
64
|
-
def reload
|
65
|
-
draws = @draws.dup
|
66
|
-
clear
|
67
|
-
draws.each { |block| draw(&block) }
|
68
|
-
end
|
69
|
-
|
70
53
|
private
|
71
54
|
|
72
55
|
# Builds and saves given consumer group
|
73
56
|
# @param group_id [String, Symbol] name for consumer group
|
74
57
|
# @param block [Proc] proc that should be executed in the proxy context
|
75
58
|
def consumer_group(group_id, &block)
|
76
|
-
consumer_group =
|
77
|
-
|
59
|
+
consumer_group = find { |cg| cg.name == group_id.to_s }
|
60
|
+
|
61
|
+
if consumer_group
|
62
|
+
Proxy.new(consumer_group, &block).target
|
63
|
+
else
|
64
|
+
consumer_group = ConsumerGroup.new(group_id.to_s)
|
65
|
+
self << Proxy.new(consumer_group, &block).target
|
66
|
+
end
|
67
|
+
end
|
68
|
+
|
69
|
+
# Handles the simple routing case where we create one consumer group and allow for further
|
70
|
+
# subscription group customization
|
71
|
+
# @param subscription_group_name [String, Symbol] subscription group id. When not provided,
|
72
|
+
# a random uuid will be used
|
73
|
+
# @param block [Proc] further topics definitions
|
74
|
+
def subscription_group(subscription_group_name = SecureRandom.uuid, &block)
|
75
|
+
consumer_group('app') do
|
76
|
+
target.public_send(:subscription_group=, subscription_group_name, &block)
|
77
|
+
end
|
78
78
|
end
|
79
79
|
|
80
|
+
# In case we use simple style of routing, all topics will be assigned to the same consumer
|
81
|
+
# group that will be based on the client_id
|
82
|
+
#
|
80
83
|
# @param topic_name [String, Symbol] name of a topic from which we want to consumer
|
81
84
|
# @param block [Proc] proc we want to evaluate in the topic context
|
82
85
|
def topic(topic_name, &block)
|
83
|
-
consumer_group(
|
84
|
-
topic(topic_name, &block)
|
86
|
+
consumer_group('app') do
|
87
|
+
topic(topic_name, &block)
|
85
88
|
end
|
86
89
|
end
|
87
90
|
end
|