[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16355288#comment-16355288
]
ASF GitHub Bot commented on FLINK-8419:
---
Github user tzulitai closed the pull request at:
https://github.com/apache/flink/pull/5336
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16354377#comment-16354377
]
Tzu-Li (Gordon) Tai commented on FLINK-8419:
Merged.
1.5 - 40f26c80476b884a6b7f3431562a78018fa014c5
1.4 - 432273aa963f46dc032c5007389a1a681b5188e6
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16354358#comment-16354358
]
ASF GitHub Bot commented on FLINK-8419:
---
Github user asfgit closed the pull request at:
https://github.com/apache/flink/pull/5335
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16346783#comment-16346783
]
ASF GitHub Bot commented on FLINK-8419:
---
Github user tzulitai commented on a diff in the pull request:
https://github.com/apache/flink/pull/5335#discussion_r165045598
--- Diff:
flink-connectors/flink-connector-kafka-0.9/src/main/java/org/apache/flink/streaming/connectors/kafka/internal/Kafka09Fetcher.java
---
@@ -95,21 +95,19 @@ public Kafka09Fetcher(
watermarksPunctuated,
processingTimeProvider,
autoWatermarkInterval,
- userCodeClassLoader,
+ userCodeClassLoader.getParent(),
+ consumerMetricGroup,
useMetrics);
this.deserializer = deserializer;
this.handover = new Handover();
- final MetricGroup kafkaMetricGroup =
metricGroup.addGroup(KAFKA_CONSUMER_METRICS_GROUP);
- addOffsetStateGauge(kafkaMetricGroup);
-
this.consumerThread = new KafkaConsumerThread(
LOG,
handover,
kafkaProperties,
unassignedPartitionsQueue,
- kafkaMetricGroup,
+ subtaskMetricGroup, // TODO: the thread should
expose Kafka-shipped metrics through the consumer metric group, not subtask
metric group
--- End diff --
Will address this as discussed in #5336, and then merge this.
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16346636#comment-16346636
]
ASF GitHub Bot commented on FLINK-8419:
---
Github user zentol commented on a diff in the pull request:
https://github.com/apache/flink/pull/5336#discussion_r165023129
--- Diff:
flink-connectors/flink-connector-kafka-0.9/src/main/java/org/apache/flink/streaming/connectors/kafka/internal/Kafka09Fetcher.java
---
@@ -92,21 +93,19 @@ public Kafka09Fetcher(
watermarksPunctuated,
processingTimeProvider,
autoWatermarkInterval,
- userCodeClassLoader,
+ userCodeClassLoader.getParent(),
+ consumerMetricGroup,
useMetrics);
this.deserializer = deserializer;
this.handover = new Handover();
- final MetricGroup kafkaMetricGroup =
metricGroup.addGroup("KafkaConsumer");
- addOffsetStateGauge(kafkaMetricGroup);
-
this.consumerThread = new KafkaConsumerThread(
LOG,
handover,
kafkaProperties,
unassignedPartitionsQueue,
- kafkaMetricGroup,
+ subtaskMetricGroup, // TODO: the thread should
expose Kafka-shipped metrics through the consumer metric group, not subtask
metric group
--- End diff --
for 1.4 I would just remove the TODO since we won't fix it, but for 1.5 I
would as you suggested register them twice.
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16346639#comment-16346639
]
ASF GitHub Bot commented on FLINK-8419:
---
Github user tzulitai commented on a diff in the pull request:
https://github.com/apache/flink/pull/5336#discussion_r165023649
--- Diff:
flink-connectors/flink-connector-kafka-0.9/src/main/java/org/apache/flink/streaming/connectors/kafka/internal/Kafka09Fetcher.java
---
@@ -92,21 +93,19 @@ public Kafka09Fetcher(
watermarksPunctuated,
processingTimeProvider,
autoWatermarkInterval,
- userCodeClassLoader,
+ userCodeClassLoader.getParent(),
+ consumerMetricGroup,
useMetrics);
this.deserializer = deserializer;
this.handover = new Handover();
- final MetricGroup kafkaMetricGroup =
metricGroup.addGroup("KafkaConsumer");
- addOffsetStateGauge(kafkaMetricGroup);
-
this.consumerThread = new KafkaConsumerThread(
LOG,
handover,
kafkaProperties,
unassignedPartitionsQueue,
- kafkaMetricGroup,
+ subtaskMetricGroup, // TODO: the thread should
expose Kafka-shipped metrics through the consumer metric group, not subtask
metric group
--- End diff --
👌 makes sense, I'll do that and merge this.
Thanks for the review!
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16346634#comment-16346634
]
ASF GitHub Bot commented on FLINK-8419:
---
Github user zentol commented on a diff in the pull request:
https://github.com/apache/flink/pull/5335#discussion_r165022709
--- Diff:
flink-connectors/flink-connector-kafka-0.9/src/main/java/org/apache/flink/streaming/connectors/kafka/internal/Kafka09Fetcher.java
---
@@ -95,21 +95,19 @@ public Kafka09Fetcher(
watermarksPunctuated,
processingTimeProvider,
autoWatermarkInterval,
- userCodeClassLoader,
+ userCodeClassLoader.getParent(),
+ consumerMetricGroup,
useMetrics);
this.deserializer = deserializer;
this.handover = new Handover();
- final MetricGroup kafkaMetricGroup =
metricGroup.addGroup(KAFKA_CONSUMER_METRICS_GROUP);
- addOffsetStateGauge(kafkaMetricGroup);
-
this.consumerThread = new KafkaConsumerThread(
LOG,
handover,
kafkaProperties,
unassignedPartitionsQueue,
- kafkaMetricGroup,
+ subtaskMetricGroup, // TODO: the thread should
expose Kafka-shipped metrics through the consumer metric group, not subtask
metric group
--- End diff --
so why aren't we passing the consumerMetricGroup here?
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16346633#comment-16346633
]
ASF GitHub Bot commented on FLINK-8419:
---
Github user tzulitai commented on a diff in the pull request:
https://github.com/apache/flink/pull/5336#discussion_r165022738
--- Diff:
flink-connectors/flink-connector-kafka-0.9/src/main/java/org/apache/flink/streaming/connectors/kafka/internal/Kafka09Fetcher.java
---
@@ -92,21 +93,19 @@ public Kafka09Fetcher(
watermarksPunctuated,
processingTimeProvider,
autoWatermarkInterval,
- userCodeClassLoader,
+ userCodeClassLoader.getParent(),
+ consumerMetricGroup,
useMetrics);
this.deserializer = deserializer;
this.handover = new Handover();
- final MetricGroup kafkaMetricGroup =
metricGroup.addGroup("KafkaConsumer");
- addOffsetStateGauge(kafkaMetricGroup);
-
this.consumerThread = new KafkaConsumerThread(
LOG,
handover,
kafkaProperties,
unassignedPartitionsQueue,
- kafkaMetricGroup,
+ subtaskMetricGroup, // TODO: the thread should
expose Kafka-shipped metrics through the consumer metric group, not subtask
metric group
--- End diff --
Because that would break compatibility of previous metrics.
On the other hand, it might make sense that we also additionally register
the Kafka-shipped metrics under `consumerMetricGroup` now, if we want to be
able to eventually resolve this TODO in the future. What do you think?
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16346624#comment-16346624
]
ASF GitHub Bot commented on FLINK-8419:
---
Github user zentol commented on a diff in the pull request:
https://github.com/apache/flink/pull/5336#discussion_r165021513
--- Diff:
flink-connectors/flink-connector-kafka-0.9/src/main/java/org/apache/flink/streaming/connectors/kafka/internal/Kafka09Fetcher.java
---
@@ -92,21 +93,19 @@ public Kafka09Fetcher(
watermarksPunctuated,
processingTimeProvider,
autoWatermarkInterval,
- userCodeClassLoader,
+ userCodeClassLoader.getParent(),
+ consumerMetricGroup,
useMetrics);
this.deserializer = deserializer;
this.handover = new Handover();
- final MetricGroup kafkaMetricGroup =
metricGroup.addGroup("KafkaConsumer");
- addOffsetStateGauge(kafkaMetricGroup);
-
this.consumerThread = new KafkaConsumerThread(
LOG,
handover,
kafkaProperties,
unassignedPartitionsQueue,
- kafkaMetricGroup,
+ subtaskMetricGroup, // TODO: the thread should
expose Kafka-shipped metrics through the consumer metric group, not subtask
metric group
--- End diff --
so why aren't we passing the consumerMetricGroup here?
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16346623#comment-16346623
]
ASF GitHub Bot commented on FLINK-8419:
---
Github user zentol commented on a diff in the pull request:
https://github.com/apache/flink/pull/5336#discussion_r165021339
--- Diff:
flink-connectors/flink-connector-kafka-base/src/main/java/org/apache/flink/streaming/connectors/kafka/internals/AbstractFetcher.java
---
@@ -560,16 +585,11 @@ private void updateMinPunctuatedWatermark(Watermark
nextWatermark) {
/**
* Add current and committed offsets to metric group.
-*
-* @param metricGroup The metric group to use
*/
- protected void addOffsetStateGauge(MetricGroup metricGroup) {
- // add current offsets to gage
- MetricGroup currentOffsets =
metricGroup.addGroup("current-offsets");
- MetricGroup committedOffsets =
metricGroup.addGroup("committed-offsets");
- for (KafkaTopicPartitionState ktp :
subscribedPartitionStates) {
- currentOffsets.gauge(ktp.getTopic() + "-" +
ktp.getPartition(), new OffsetGauge(ktp, OffsetGaugeType.CURRENT_OFFSET));
- committedOffsets.gauge(ktp.getTopic() + "-" +
ktp.getPartition(), new OffsetGauge(ktp, OffsetGaugeType.COMMITTED_OFFSET));
+ protected void
registerOffsetMetrics(List>
partitionOffsetStates) {
--- End diff --
make private?
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16334420#comment-16334420
]
ASF GitHub Bot commented on FLINK-8419:
---
GitHub user tzulitai opened a pull request:
https://github.com/apache/flink/pull/5336
(release-1.4) [FLINK-8419] [kafka] Register metrics for dynamically
discovered Kafka partitions
## What is the purpose of the change
Different version of #5335, which is targeted for `release-1.4`.
This version does not include the new offset metrics added in #5214 for the
new partitions (those new metrics are only added in `master`).
## Brief change log
- db39ec2: Preliminary cleanup of the registration of the `KafkaConsumer`
user scope metric group. This commit refactors that for better separation of
concerns and lessen code duplication.
- 1acfc15: Register offset metrics for new partitions in
`addDiscoveredPartitions`
## Verifying this change
No new tests were added.
By manually running a Flink job using the Kafka consumer and repartitioning
the Kafka topic, you should be able to see metrics for the newly added
partitions.
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): (yes / **no**)
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: (yes / **no**)
- The serializers: (yes / **no** / don't know)
- The runtime per-record code paths (performance sensitive): (yes /
**no** / don't know)
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Yarn/Mesos, ZooKeeper: (yes / **no** / don't know)
- The S3 file system connector: (yes / **no** / don't know)
## Documentation
- Does this pull request introduce a new feature? (yes / **no**)
- If yes, how is the feature documented? (**not applicable** / docs /
JavaDocs / not documented)
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/tzulitai/flink FLINK-8419-1.4
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/5336.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #5336
commit db39ec2ed6f6b9cf46bc851125b859501d9a9df0
Author: Tzu-Li (Gordon) Tai
Date: 2018-01-22T12:36:20Z
[FLINK-8419] [kafka] Move consumer metric group registration to
FlinkKafkaConsumerBase
This commit is a refactor to move the registration of the consumer
metric group (user scope "KafkaConsumer") to FlinkKafkaConsumerBase.
Previously, the registration was scattered around in Kafka
version-specific subclasses.
commit 1acfc1526e56abd5e5cb4f5a32641afbf2282905
Author: Tzu-Li (Gordon) Tai
Date: 2018-01-22T13:14:44Z
[FLINK-8419] [kafka] Register metrics for dynamically discovered Kafka
partitions
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (FLINK-8419) Kafka consumer's offset metrics are not registered for dynamically discovered partitions
[
https://issues.apache.org/jira/browse/FLINK-8419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16334385#comment-16334385
]
ASF GitHub Bot commented on FLINK-8419:
---
GitHub user tzulitai opened a pull request:
https://github.com/apache/flink/pull/5335
(master) [FLINK-8419] [kafka] Register metrics for dynamically discovered
Kafka partitions
## What is the purpose of the change
This PR fixes that offset metrics (i.e. current offset, committed offsets)
were not registered for partitions that were dynamically discovered in the
`FlinkKafkaConsumerBase`.
This version is targeted for merge to `master`.
Another version targeted for `release-1.4`, which does not include the new
offset metrics added in #5214, will be separately opened.
## Brief change log
- 54f3cfd: Preliminary cleanup of the registration of the `KafkaConsumer`
user scope metric group. This commit refactors that for better separation of
concerns and lessen code duplication.
- bf1e4ce: Register offset metrics for new partitions in
`addDiscoveredPartitions`
- 7e90a75: Minor hotfix to inappropriate access modifiers in the
`AbstractFetcher`
## Verifying this change
No new tests were added.
By manually running a Flink job using the Kafka consumer and repartitioning
the Kafka topic, you should be able to see metrics for the newly added
partitions.
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): (yes / **no**)
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: (yes / **no**)
- The serializers: (yes / **no** / don't know)
- The runtime per-record code paths (performance sensitive): (yes /
**no** / don't know)
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Yarn/Mesos, ZooKeeper: (yes / **no** / don't know)
- The S3 file system connector: (yes / **no** / don't know)
## Documentation
- Does this pull request introduce a new feature? (yes / **no**)
- If yes, how is the feature documented? (**not applicable** / docs /
JavaDocs / not documented)
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/tzulitai/flink FLINK-8419
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/5335.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #5335
commit 54f3cfd3eb9f925266de37219ab56703a26795d6
Author: Tzu-Li (Gordon) Tai
Date: 2018-01-22T12:36:20Z
[FLINK-8419] [kafka] Move consumer metric group registration to
FlinkKafkaConsumerBase
This commit is a refactor to move the registration of the consumer
metric group (user scope "KafkaConsumer") to FlinkKafkaConsumerBase.
Previously, the registration was scattered around in Kafka
version-specific subclasses.
commit bf1e4ce73279a90d625cda66ee84c194d5ce3e34
Author: Tzu-Li (Gordon) Tai
Date: 2018-01-22T13:14:44Z
[FLINK-8419] [kafka] Register metrics for dynamically discovered Kafka
partitions
commit 7e90a753441b7a4af1dd400942d88fba7d0178df
Author: Tzu-Li (Gordon) Tai
Date: 2018-01-22T13:17:54Z
[hotfix] [kafka] Fix inapproriate access modifiers in AbstractFetcher
> Kafka consumer's offset metrics are not registered for dynamically discovered
> partitions
>
>
> Key: FLINK-8419
> URL: https://issues.apache.org/jira/browse/FLINK-8419
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector, Metrics
>Affects Versions: 1.4.0, 1.5.0
>Reporter: Tzu-Li (Gordon) Tai
>Assignee: Tzu-Li (Gordon) Tai
>Priority: Blocker
> Fix For: 1.5.0, 1.4.1
>
>
> Currently, the per-partition offset metrics are registered via the
> {{AbstractFetcher#addOffsetStateGauge}} method. That method is only ever
> called for the initial startup partitions, and not for dynamically discovered
> partitions.
> We should consider adding some unit tests to make sure that metrics are
> properly registered for all partitions. That would also safeguard us from
> accidentally removing metrics.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
