karafka 2.1.6 → 2.1.8
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/CHANGELOG.md +16 -1
- data/Gemfile.lock +12 -12
- data/lib/karafka/admin.rb +2 -7
- data/lib/karafka/base_consumer.rb +10 -0
- data/lib/karafka/connection/client.rb +23 -1
- data/lib/karafka/connection/rebalance_manager.rb +11 -0
- data/lib/karafka/instrumentation/logger_listener.rb +3 -0
- data/lib/karafka/messages/messages.rb +4 -2
- data/lib/karafka/pro/iterator/tpl_builder.rb +18 -3
- data/lib/karafka/pro/iterator.rb +7 -5
- data/lib/karafka/processing/executor.rb +1 -1
- data/lib/karafka/routing/topic.rb +2 -1
- data/lib/karafka/setup/attributes_map.rb +2 -0
- data/lib/karafka/version.rb +1 -1
- data.tar.gz.sig +0 -0
- metadata +2 -2
- metadata.gz.sig +0 -0
checksums.yaml
CHANGED
@@ -1,7 +1,7 @@
|
|
1
1
|
---
|
2
2
|
SHA256:
|
3
|
-
metadata.gz:
|
4
|
-
data.tar.gz:
|
3
|
+
metadata.gz: 042f365fb134a24ae360d678590ce798014751c8a23fb267001c920a42aa5324
|
4
|
+
data.tar.gz: ba2950de557a5f6c775577ce392d60ee839184f50b7d9225969c684625c9ecd0
|
5
5
|
SHA512:
|
6
|
-
metadata.gz:
|
7
|
-
data.tar.gz:
|
6
|
+
metadata.gz: 30b5fcd92c348c50482cb84542380ad28b317d47e01efad2d2049cd3ba7872c5f66a7a4fde93d8bfa262d7fcb3745dedbdb89a4fd5e6cdf2288a43606ea2361d
|
7
|
+
data.tar.gz: d9c8ba95b2b71f46a3d35e2f5be634473a7aa9f45d9b5924e1019dcb53c7cea6aca8c594f9f4d3bf85a14176c3cb98a7167b4eb6e1f6f2192059d8040440e4a9
|
checksums.yaml.gz.sig
CHANGED
Binary file
|
data/CHANGELOG.md
CHANGED
@@ -1,5 +1,20 @@
|
|
1
1
|
# Karafka framework changelog
|
2
2
|
|
3
|
+
## 2.1.8 (2023-07-29)
|
4
|
+
- [Improvement] Introduce `Karafka::BaseConsumer#used?` method to indicate, that at least one invocation of `#consume` took or will take place. This can be used as a replacement to the non-direct `messages.count` check for shutdown and revocation to ensure, that the consumption took place or is taking place (in case of running LRJ).
|
5
|
+
- [Improvement] Make `messages#to_a` return copy of the underlying array to prevent scenarios, where the mutation impacts offset management.
|
6
|
+
- [Improvement] Mitigate a librdkafka `cooperative-sticky` rebalance crash issue.
|
7
|
+
- [Improvement] Provide ability to overwrite `consumer_persistence` per subscribed topic. This is mostly useful for plugins and extensions developers.
|
8
|
+
- [Fix] Fix a case where the performance tracker would crash in case of mutation of messages to an empty state.
|
9
|
+
|
10
|
+
## 2.1.7 (2023-07-22)
|
11
|
+
- [Improvement] Always query for watermarks in the Iterator to improve the initial response time.
|
12
|
+
- [Improvement] Add `max_wait_time` option to the Iterator.
|
13
|
+
- [Fix] Fix a case where `Admin#read_topic` would wait for poll interval on non-existing messages instead of early exit.
|
14
|
+
- [Fix] Fix a case where Iterator with per partition offsets with negative lookups would go below the number of available messages.
|
15
|
+
- [Fix] Remove unused constant from Admin module.
|
16
|
+
- [Fix] Add missing `connection.client.rebalance_callback.error` to the `LoggerListener` instrumentation hook.
|
17
|
+
|
3
18
|
## 2.1.6 (2023-06-29)
|
4
19
|
- [Improvement] Provide time support for iterator
|
5
20
|
- [Improvement] Provide time support for admin `#read_topic`
|
@@ -63,7 +78,7 @@
|
|
63
78
|
2. Replace `Karafka::Pro::BaseConsumer` references to `Karafka::BaseConsumer`.
|
64
79
|
3. Replace `Karafka::Instrumentation::Vendors::Datadog:Listener` with `Karafka::Instrumentation::Vendors::Datadog::MetricsListener`.
|
65
80
|
|
66
|
-
## 2.0.41 (2023-
|
81
|
+
## 2.0.41 (2023-04-19)
|
67
82
|
- **[Feature]** Provide `Karafka::Pro::Iterator` for anonymous topic/partitions iterations and messages lookups (#1389 and #1427).
|
68
83
|
- [Improvement] Optimize topic lookup for `read_topic` admin method usage.
|
69
84
|
- [Improvement] Report via `LoggerListener` information about the partition on which a given job has started and finished.
|
data/Gemfile.lock
CHANGED
@@ -1,7 +1,7 @@
|
|
1
1
|
PATH
|
2
2
|
remote: .
|
3
3
|
specs:
|
4
|
-
karafka (2.1.
|
4
|
+
karafka (2.1.8)
|
5
5
|
karafka-core (>= 2.1.1, < 2.2.0)
|
6
6
|
thor (>= 0.20)
|
7
7
|
waterdrop (>= 2.6.2, < 3.0.0)
|
@@ -10,10 +10,10 @@ PATH
|
|
10
10
|
GEM
|
11
11
|
remote: https://rubygems.org/
|
12
12
|
specs:
|
13
|
-
activejob (7.0.
|
14
|
-
activesupport (= 7.0.
|
13
|
+
activejob (7.0.6)
|
14
|
+
activesupport (= 7.0.6)
|
15
15
|
globalid (>= 0.3.6)
|
16
|
-
activesupport (7.0.
|
16
|
+
activesupport (7.0.6)
|
17
17
|
concurrent-ruby (~> 1.0, >= 1.0.2)
|
18
18
|
i18n (>= 1.6, < 2)
|
19
19
|
minitest (>= 5.1)
|
@@ -33,21 +33,21 @@ GEM
|
|
33
33
|
karafka-core (2.1.1)
|
34
34
|
concurrent-ruby (>= 1.1)
|
35
35
|
karafka-rdkafka (>= 0.13.1, < 0.14.0)
|
36
|
-
karafka-rdkafka (0.13.
|
36
|
+
karafka-rdkafka (0.13.3)
|
37
37
|
ffi (~> 1.15)
|
38
38
|
mini_portile2 (~> 2.6)
|
39
39
|
rake (> 12)
|
40
|
-
karafka-web (0.6.
|
40
|
+
karafka-web (0.6.3)
|
41
41
|
erubi (~> 1.4)
|
42
42
|
karafka (>= 2.1.4, < 3.0.0)
|
43
43
|
karafka-core (>= 2.0.13, < 3.0.0)
|
44
44
|
roda (~> 3.68, >= 3.68)
|
45
45
|
tilt (~> 2.0)
|
46
|
-
mini_portile2 (2.8.
|
46
|
+
mini_portile2 (2.8.4)
|
47
47
|
minitest (5.18.1)
|
48
48
|
rack (3.0.8)
|
49
49
|
rake (13.0.6)
|
50
|
-
roda (3.
|
50
|
+
roda (3.70.0)
|
51
51
|
rack
|
52
52
|
rspec (3.12.0)
|
53
53
|
rspec-core (~> 3.12.0)
|
@@ -58,10 +58,10 @@ GEM
|
|
58
58
|
rspec-expectations (3.12.3)
|
59
59
|
diff-lcs (>= 1.2.0, < 2.0)
|
60
60
|
rspec-support (~> 3.12.0)
|
61
|
-
rspec-mocks (3.12.
|
61
|
+
rspec-mocks (3.12.6)
|
62
62
|
diff-lcs (>= 1.2.0, < 2.0)
|
63
63
|
rspec-support (~> 3.12.0)
|
64
|
-
rspec-support (3.12.
|
64
|
+
rspec-support (3.12.1)
|
65
65
|
simplecov (0.22.0)
|
66
66
|
docile (~> 1.1)
|
67
67
|
simplecov-html (~> 0.11)
|
@@ -72,8 +72,8 @@ GEM
|
|
72
72
|
tilt (2.2.0)
|
73
73
|
tzinfo (2.0.6)
|
74
74
|
concurrent-ruby (~> 1.0)
|
75
|
-
waterdrop (2.6.
|
76
|
-
karafka-core (>= 2.1.
|
75
|
+
waterdrop (2.6.5)
|
76
|
+
karafka-core (>= 2.1.1, < 3.0.0)
|
77
77
|
zeitwerk (~> 2.3)
|
78
78
|
zeitwerk (2.6.8)
|
79
79
|
|
data/lib/karafka/admin.rb
CHANGED
@@ -9,11 +9,6 @@ module Karafka
|
|
9
9
|
# @note It always uses the primary defined cluster and does not support multi-cluster work.
|
10
10
|
# If you need this, just replace the cluster info for the time you use this
|
11
11
|
module Admin
|
12
|
-
# A fake admin topic representation that we use for messages fetched using this API
|
13
|
-
# We cannot use the topics directly because we may want to request data from topics that we
|
14
|
-
# do not have in the routing
|
15
|
-
Topic = Struct.new(:name, :deserializer)
|
16
|
-
|
17
12
|
# We wait only for this amount of time before raising error as we intercept this error and
|
18
13
|
# retry after checking that the operation was finished or failed using external factor.
|
19
14
|
MAX_WAIT_TIMEOUT = 1
|
@@ -37,7 +32,7 @@ module Karafka
|
|
37
32
|
'enable.auto.commit': false
|
38
33
|
}.freeze
|
39
34
|
|
40
|
-
private_constant :
|
35
|
+
private_constant :CONFIG_DEFAULTS, :MAX_WAIT_TIMEOUT, :TPL_REQUEST_TIMEOUT,
|
41
36
|
:MAX_ATTEMPTS
|
42
37
|
|
43
38
|
class << self
|
@@ -71,7 +66,7 @@ module Karafka
|
|
71
66
|
requested_range = (start_offset..start_offset + (count - 1))
|
72
67
|
# Establish theoretical available range. Note, that this does not handle cases related to
|
73
68
|
# log retention or compaction
|
74
|
-
available_range = (low_offset..high_offset)
|
69
|
+
available_range = (low_offset..(high_offset - 1))
|
75
70
|
# Select only offset that we can select. This will remove all the potential offsets that
|
76
71
|
# are below the low watermark offset
|
77
72
|
possible_range = requested_range.select { |offset| available_range.include?(offset) }
|
@@ -25,6 +25,7 @@ module Karafka
|
|
25
25
|
# Creates new consumer and assigns it an id
|
26
26
|
def initialize
|
27
27
|
@id = SecureRandom.hex(6)
|
28
|
+
@used = false
|
28
29
|
end
|
29
30
|
|
30
31
|
# Can be used to run preparation code prior to the job being enqueued
|
@@ -34,6 +35,7 @@ module Karafka
|
|
34
35
|
# not as a part of the public api. This should not perform any extensive operations as it is
|
35
36
|
# blocking and running in the listener thread.
|
36
37
|
def on_before_enqueue
|
38
|
+
@used = true
|
37
39
|
handle_before_enqueue
|
38
40
|
rescue StandardError => e
|
39
41
|
Karafka.monitor.instrument(
|
@@ -160,6 +162,14 @@ module Karafka
|
|
160
162
|
# some teardown procedures (closing file handler, etc).
|
161
163
|
def shutdown; end
|
162
164
|
|
165
|
+
# @return [Boolean] was this consumer in active use. Active use means running `#consume` at
|
166
|
+
# least once. Consumer may have to run `#revoked` or `#shutdown` despite not running
|
167
|
+
# `#consume` previously in delayed job cases and other cases that potentially involve running
|
168
|
+
# the `Jobs::Idle` for house-keeping
|
169
|
+
def used?
|
170
|
+
@used
|
171
|
+
end
|
172
|
+
|
163
173
|
# Pauses processing on a given offset for the current topic partition
|
164
174
|
#
|
165
175
|
# After given partition is resumed, it will continue processing from the given offset
|
@@ -23,11 +23,17 @@ module Karafka
|
|
23
23
|
# Max time for a TPL request. We increase it to compensate for remote clusters latency
|
24
24
|
TPL_REQUEST_TIMEOUT = 2_000
|
25
25
|
|
26
|
+
# 1 minute of max wait for the first rebalance before a forceful attempt
|
27
|
+
# This applies only to a case when a short-lived Karafka instance with a client would be
|
28
|
+
# closed before first rebalance. Mitigates a librdkafka bug.
|
29
|
+
COOPERATIVE_STICKY_MAX_WAIT = 60_000
|
30
|
+
|
26
31
|
# We want to make sure we never close several clients in the same moment to prevent
|
27
32
|
# potential race conditions and other issues
|
28
33
|
SHUTDOWN_MUTEX = Mutex.new
|
29
34
|
|
30
|
-
private_constant :MAX_POLL_RETRIES, :SHUTDOWN_MUTEX, :TPL_REQUEST_TIMEOUT
|
35
|
+
private_constant :MAX_POLL_RETRIES, :SHUTDOWN_MUTEX, :TPL_REQUEST_TIMEOUT,
|
36
|
+
:COOPERATIVE_STICKY_MAX_WAIT
|
31
37
|
|
32
38
|
# Creates a new consumer instance.
|
33
39
|
#
|
@@ -226,6 +232,22 @@ module Karafka
|
|
226
232
|
# as until all the consumers are stopped, the server will keep running serving only
|
227
233
|
# part of the messages
|
228
234
|
def stop
|
235
|
+
# This ensures, that we do not stop the underlying client until it passes the first
|
236
|
+
# rebalance for cooperative-sticky. Otherwise librdkafka may crash
|
237
|
+
#
|
238
|
+
# We set a timeout just in case the rebalance would never happen or would last for an
|
239
|
+
# extensive time period.
|
240
|
+
#
|
241
|
+
# @see https://github.com/confluentinc/librdkafka/issues/4312
|
242
|
+
if @subscription_group.kafka[:'partition.assignment.strategy'] == 'cooperative-sticky'
|
243
|
+
(COOPERATIVE_STICKY_MAX_WAIT / 100).times do
|
244
|
+
# If we're past the first rebalance, no need to wait
|
245
|
+
break if @rebalance_manager.active?
|
246
|
+
|
247
|
+
sleep(0.1)
|
248
|
+
end
|
249
|
+
end
|
250
|
+
|
229
251
|
close
|
230
252
|
end
|
231
253
|
|
@@ -30,6 +30,7 @@ module Karafka
|
|
30
30
|
@assigned_partitions = {}
|
31
31
|
@revoked_partitions = {}
|
32
32
|
@changed = false
|
33
|
+
@active = false
|
33
34
|
end
|
34
35
|
|
35
36
|
# Resets the rebalance manager state
|
@@ -46,11 +47,20 @@ module Karafka
|
|
46
47
|
@changed
|
47
48
|
end
|
48
49
|
|
50
|
+
# @return [Boolean] true if there was at least one rebalance
|
51
|
+
# @note This method is needed to make sure that when using cooperative-sticky, we do not
|
52
|
+
# close until first rebalance. Otherwise librdkafka may crash.
|
53
|
+
# @see https://github.com/confluentinc/librdkafka/issues/4312
|
54
|
+
def active?
|
55
|
+
@active
|
56
|
+
end
|
57
|
+
|
49
58
|
# Callback that kicks in inside of rdkafka, when new partitions are assigned.
|
50
59
|
#
|
51
60
|
# @private
|
52
61
|
# @param partitions [Rdkafka::Consumer::TopicPartitionList]
|
53
62
|
def on_partitions_assigned(partitions)
|
63
|
+
@active = true
|
54
64
|
@assigned_partitions = partitions.to_h.transform_values { |part| part.map(&:partition) }
|
55
65
|
@changed = true
|
56
66
|
end
|
@@ -60,6 +70,7 @@ module Karafka
|
|
60
70
|
# @private
|
61
71
|
# @param partitions [Rdkafka::Consumer::TopicPartitionList]
|
62
72
|
def on_partitions_revoked(partitions)
|
73
|
+
@active = true
|
63
74
|
@revoked_partitions = partitions.to_h.transform_values { |part| part.map(&:partition) }
|
64
75
|
@changed = true
|
65
76
|
end
|
@@ -277,6 +277,9 @@ module Karafka
|
|
277
277
|
when 'connection.client.poll.error'
|
278
278
|
error "Data polling error occurred: #{error}"
|
279
279
|
error details
|
280
|
+
when 'connection.client.rebalance_callback.error'
|
281
|
+
error "Rebalance callback error occurred: #{error}"
|
282
|
+
error details
|
280
283
|
else
|
281
284
|
# This should never happen. Please contact the maintainers
|
282
285
|
raise Errors::UnsupportedCaseError, event
|
@@ -60,10 +60,12 @@ module Karafka
|
|
60
60
|
@messages_array.size
|
61
61
|
end
|
62
62
|
|
63
|
-
# @return [Array<Karafka::Messages::Message>] pure array with messages
|
63
|
+
# @return [Array<Karafka::Messages::Message>] copy of the pure array with messages
|
64
64
|
def to_a
|
65
|
-
@messages_array
|
65
|
+
@messages_array.dup
|
66
66
|
end
|
67
|
+
|
68
|
+
alias count size
|
67
69
|
end
|
68
70
|
end
|
69
71
|
end
|
@@ -93,12 +93,27 @@ module Karafka
|
|
93
93
|
next unless partitions.is_a?(Hash)
|
94
94
|
|
95
95
|
partitions.each do |partition, offset|
|
96
|
+
# Care only about numerical offsets
|
97
|
+
#
|
98
|
+
# For time based we already resolve them via librdkafka lookup API
|
99
|
+
next unless offset.is_a?(Integer)
|
100
|
+
|
101
|
+
low_offset, high_offset = @consumer.query_watermark_offsets(name, partition)
|
102
|
+
|
96
103
|
# Care only about negative offsets (last n messages)
|
97
|
-
|
104
|
+
#
|
105
|
+
# We reject the above results but we **NEED** to run the `#query_watermark_offsets`
|
106
|
+
# for each topic partition nonetheless. Without this, librdkafka fetches a lot more
|
107
|
+
# metadata about each topic and each partition and this takes much more time than
|
108
|
+
# just getting watermarks. If we do not run watermark, at least an extra second
|
109
|
+
# is added at the beginning of iterator flow
|
110
|
+
#
|
111
|
+
# This may not be significant when this runs in the background but in case of
|
112
|
+
# using iterator in thins like Puma, it heavily impacts the end user experience
|
113
|
+
next unless offset.negative?
|
98
114
|
|
99
|
-
_, high_watermark_offset = @consumer.query_watermark_offsets(name, partition)
|
100
115
|
# We add because this offset is negative
|
101
|
-
@mapped_topics[name][partition] =
|
116
|
+
@mapped_topics[name][partition] = [high_offset + offset, low_offset].max
|
102
117
|
end
|
103
118
|
end
|
104
119
|
end
|
data/lib/karafka/pro/iterator.rb
CHANGED
@@ -39,6 +39,7 @@ module Karafka
|
|
39
39
|
# overwritten, you may want to include `auto.offset.reset` to match your case.
|
40
40
|
# @param yield_nil [Boolean] should we yield also `nil` values when poll returns nothing.
|
41
41
|
# Useful in particular for long-living iterators.
|
42
|
+
# @param max_wait_time [Integer] max wait in ms when iterator did not receive any messages
|
42
43
|
#
|
43
44
|
# @note It is worth keeping in mind, that this API also needs to operate within
|
44
45
|
# `max.poll.interval.ms` limitations on each iteration
|
@@ -48,7 +49,8 @@ module Karafka
|
|
48
49
|
def initialize(
|
49
50
|
topics,
|
50
51
|
settings: { 'auto.offset.reset': 'beginning' },
|
51
|
-
yield_nil: false
|
52
|
+
yield_nil: false,
|
53
|
+
max_wait_time: 200
|
52
54
|
)
|
53
55
|
@topics_with_partitions = Expander.new.call(topics)
|
54
56
|
|
@@ -62,6 +64,7 @@ module Karafka
|
|
62
64
|
|
63
65
|
@settings = settings
|
64
66
|
@yield_nil = yield_nil
|
67
|
+
@max_wait_time = max_wait_time
|
65
68
|
end
|
66
69
|
|
67
70
|
# Iterates over requested topic partitions and yields the results with the iterator itself
|
@@ -80,7 +83,7 @@ module Karafka
|
|
80
83
|
# Stream data until we reach the end of all the partitions or until the end user
|
81
84
|
# indicates that they are done
|
82
85
|
until done?
|
83
|
-
message = poll
|
86
|
+
message = poll
|
84
87
|
|
85
88
|
# Skip nils if not explicitly required
|
86
89
|
next if message.nil? && !@yield_nil
|
@@ -131,10 +134,9 @@ module Karafka
|
|
131
134
|
|
132
135
|
private
|
133
136
|
|
134
|
-
# @param timeout [Integer] timeout in ms
|
135
137
|
# @return [Rdkafka::Consumer::Message, nil] message or nil if nothing to do
|
136
|
-
def poll
|
137
|
-
@current_consumer.poll(
|
138
|
+
def poll
|
139
|
+
@current_consumer.poll(@max_wait_time)
|
138
140
|
rescue Rdkafka::RdkafkaError => e
|
139
141
|
# End of partition
|
140
142
|
if e.code == :partition_eof
|
@@ -48,7 +48,7 @@ module Karafka
|
|
48
48
|
# We reload the consumers with each batch instead of relying on some external signals
|
49
49
|
# when needed for consistency. That way devs may have it on or off and not in this
|
50
50
|
# middle state, where re-creation of a consumer instance would occur only sometimes
|
51
|
-
@consumer = nil unless
|
51
|
+
@consumer = nil unless topic.consumer_persistence
|
52
52
|
|
53
53
|
# First we build messages batch...
|
54
54
|
consumer.messages = Messages::Builders::Messages.call(
|
@@ -17,6 +17,7 @@ module Karafka
|
|
17
17
|
max_messages
|
18
18
|
max_wait_time
|
19
19
|
initial_offset
|
20
|
+
consumer_persistence
|
20
21
|
].freeze
|
21
22
|
|
22
23
|
private_constant :INHERITABLE_ATTRIBUTES
|
@@ -50,7 +51,7 @@ module Karafka
|
|
50
51
|
|
51
52
|
# @return [Class] consumer class that we should use
|
52
53
|
def consumer
|
53
|
-
if
|
54
|
+
if consumer_persistence
|
54
55
|
# When persistence of consumers is on, no need to reload them
|
55
56
|
@consumer
|
56
57
|
else
|
@@ -25,6 +25,7 @@ module Karafka
|
|
25
25
|
broker.version.fallback
|
26
26
|
builtin.features
|
27
27
|
check.crcs
|
28
|
+
client.dns.lookup
|
28
29
|
client.id
|
29
30
|
client.rack
|
30
31
|
closesocket_cb
|
@@ -161,6 +162,7 @@ module Karafka
|
|
161
162
|
broker.address.ttl
|
162
163
|
broker.version.fallback
|
163
164
|
builtin.features
|
165
|
+
client.dns.lookup
|
164
166
|
client.id
|
165
167
|
client.rack
|
166
168
|
closesocket_cb
|
data/lib/karafka/version.rb
CHANGED
data.tar.gz.sig
CHANGED
Binary file
|
metadata
CHANGED
@@ -1,7 +1,7 @@
|
|
1
1
|
--- !ruby/object:Gem::Specification
|
2
2
|
name: karafka
|
3
3
|
version: !ruby/object:Gem::Version
|
4
|
-
version: 2.1.
|
4
|
+
version: 2.1.8
|
5
5
|
platform: ruby
|
6
6
|
authors:
|
7
7
|
- Maciej Mensfeld
|
@@ -35,7 +35,7 @@ cert_chain:
|
|
35
35
|
Qf04B9ceLUaC4fPVEz10FyobjaFoY4i32xRto3XnrzeAgfEe4swLq8bQsR3w/EF3
|
36
36
|
MGU0FeSV2Yj7Xc2x/7BzLK8xQn5l7Yy75iPF+KP3vVmDHnNl
|
37
37
|
-----END CERTIFICATE-----
|
38
|
-
date: 2023-
|
38
|
+
date: 2023-07-29 00:00:00.000000000 Z
|
39
39
|
dependencies:
|
40
40
|
- !ruby/object:Gem::Dependency
|
41
41
|
name: karafka-core
|
metadata.gz.sig
CHANGED
Binary file
|