[
https://issues.apache.org/jira/browse/FLINK-4022?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15896986#comment-15896986
]
ASF GitHub Bot commented on FLINK-4022:
---------------------------------------
GitHub user tzulitai opened a pull request:
https://github.com/apache/flink/pull/3476
[FLINK-4022] [kafka] Partition and topic pattern discovery for
FlinkKafkaConsumer
This PR adds partition and topic regex pattern discovery to the
`FlinkKafkaConsumer`.
It doesn't expose a new constructor that accepts regex topic patterns yet.
I propose to expose that with https://issues.apache.org/jira/browse/FLINK-5704
(deprecate the original `FlinkKafkaConsumer` constructors in favor of new ones
with new offset behaviours). For this reason, I also propose to update the
Kafka documentation when the new constructors are added.
### Design
Some description to ease review:
- `AbstractPartitionDiscoverer`:
A `AbstractPartitionDiscoverer` is a stateful utility instance that
remembers what partitions are discovered already. It also wraps the logic for
partition-to-subtask assignment. The main `run()` method now has a discovery
loop that calls `AbstractPartitionDiscoverer#discoverPartitions()` on a fixed
interval. This method returns only new partitions that should be subscribed by
the subtask.
The returned partitions are used to invoke
`AbstractFetcher#addDiscoveredPartitions(...)` on the fetcher.
On a fresh startup, `AbstractPartitionDiscoverer#discoverPartitions()` is
also used to fetch the initial seed startup partitions in `open()`.
- `AbstractFetcher#addDiscoveredPartitions(...)`
The fetcher now has a `unassignedPartitionsQueue` that contains discovered
partitions not yet consumed by concrete Kafka clients. Whenever
`addDiscoveredPartitions(...)` is called on the fetcher, the fetcher will
create the state holders for the partitions, and add the partitions to the
queue.
Concrete implementations of the fetcher should continuously poll this queue
in the fetch loop. If partitions are found from the queue, they should be
assigned for consuming.
- Concrete fetchers continuously polls the queue in `runFetchLoop()`
For 0.8, this simply means that the original `unassignedPartitionsQueue` in
`Kafka08Fetcher` is moved to the base abstract fetcher class. Nothing else is
touched.
For 0.9+, queue polling and partition reassignment for the high-level
consumer happens in `KafkaConsumerThread`.
### TODOs
This PR serves as a preview for the new functionality. Below are some
pending TODOs
- Currently, partition discovery will not work correctly after restore. The
reason for this is explained with `TODO` comments within the
`FlinkKafkaConsumerBase#open()` method. For this to work correctly, @rmetzger
and I are considering 2 options: 1) use broadcast state, or 2) assign
partitions using `maxParallelism` and `assignedKeyGroupIds` instead of subtask
index / number of subtasks.
- The PR still lacks exactly-once integration tests with Kafka
repartitioning / dynamic topics.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/tzulitai/flink FLINK-4022
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/3476.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 #3476
----
commit 469a3c964e13ff559afd1d46933113bcff48eafc
Author: Tzu-Li (Gordon) Tai <[email protected]>
Date: 2017-03-06T09:20:32Z
[FLINK-4022] [kafka] Partition and topic pattern discovery for
FlinkKafkaConsumer
----
> Partition discovery / regex topic subscription for the Kafka consumer
> ---------------------------------------------------------------------
>
> Key: FLINK-4022
> URL: https://issues.apache.org/jira/browse/FLINK-4022
> Project: Flink
> Issue Type: New Feature
> Components: Kafka Connector, Streaming Connectors
> Affects Versions: 1.0.0
> Reporter: Tzu-Li (Gordon) Tai
> Assignee: Tzu-Li (Gordon) Tai
>
> Example: allow users to subscribe to "topic-n*", so that the consumer
> automatically reads from "topic-n1", "topic-n2", ... and so on as they are
> added to Kafka.
> I propose to implement this feature by the following description:
> Since the overall list of partitions to read will change after job
> submission, the main big change required for this feature will be dynamic
> partition assignment to subtasks while the Kafka consumer is running. This
> will mainly be accomplished using Kafka 0.9.x API
> `KafkaConsumer#subscribe(java.util.regex.Pattern,
> ConsumerRebalanceListener)`. Each KafkaConsumers in each subtask will be
> added to the same consumer group when instantiated, and rely on Kafka to
> dynamically reassign partitions to them whenever a rebalance happens. The
> registered `ConsumerRebalanceListener` is a callback that is called right
> before and after rebalancing happens. We'll use this callback to let each
> subtask commit its last offsets of partitions its currently responsible of to
> an external store (or Kafka) before a rebalance; after rebalance and the
> substasks gets the new partitions it'll be reading from, they'll read from
> the external store to get the last offsets for their new partitions
> (partitions which don't have offset entries in the store are new partitions
> causing the rebalancing).
> The tricky part will be restoring Flink checkpoints when the partition
> assignment is dynamic. Snapshotting will remain the same - subtasks snapshot
> the offsets of partitions they are currently holding. Restoring will be a
> bit different in that subtasks might not be assigned matching partitions to
> the snapshot the subtask is restored with (since we're letting Kafka
> dynamically assign partitions). There will need to be a coordination process
> where, if a restore state exists, all subtasks first commit the offsets they
> receive (as a result of the restore state) to the external store, and then
> all subtasks attempt to find a last offset for the partitions it is holding.
> However, if the globally merged restore state feature mentioned by
> [~StephanEwen] in https://issues.apache.org/jira/browse/FLINK-3231 is
> available, then the restore will be simple again, as each subtask has full
> access to previous global state therefore coordination is not required.
> I think changing to dynamic partition assignment is also good in the long run
> for handling topic repartitioning.
> Overall,
> User-facing API changes:
> - New constructor - FlinkKafkaConsumer09(java.util.regex.Pattern,
> DeserializationSchema, Properties)
> - New constructor - FlinkKafkaConsumer09(java.util.regex.Pattern,
> KeyedDeserializationSchema, Properties)
> Implementation changes:
> 1. Dynamic partition assigning depending on KafkaConsumer#subscribe
> - Remove partition list querying from constructor
> - Remove static partition assigning to substasks in run()
> - Instead of using KafkaConsumer#assign() in fetchers to manually assign
> static partitions, use subscribe() registered with the callback
> implementation explained above.
> 2. Restoring from checkpointed states
> - Snapshotting should remain unchanged
> - Restoring requires subtasks to coordinate the restored offsets they hold
> before continuing (unless we are able to have merged restore states).
> 3. For previous consumer functionality (consume from fixed list of topics),
> the KafkaConsumer#subscribe() has a corresponding overload method for fixed
> list of topics. We can simply decide which subscribe() overload to use
> depending on whether a regex Pattern or list of topics is supplied.
> 4. If subtasks don't initially have any assigned partitions, we shouldn't
> emit MAX_VALUE watermark, since it may hold partitions after a rebalance.
> Instead, un-assigned subtasks should be running a fetcher instance too and
> take part as a process pool for the consumer group of the subscribed topics.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
