The backing store for offsets, status, and configs now attempts to use the new AdminClient to look up the internal topics and create them if they don’t yet exist. If the necessary APIs are not available in the connected broker, the stores fall back to the old behavior of relying upon auto-created topics. Kafka Connect requires a minimum of Apache Kafka 0.10.0.1-cp1, and the AdminClient can work with all versions since 0.10.0.0.
All three of Connect’s internal topics are created as compacted topics, and new distributed worker configuration properties control the replication factor for all three topics and the number of partitions for the offsets and status topics; the config topic requires a single partition and does not allow it to be set via configuration. All of these new configuration properties have sensible defaults, meaning users can upgrade without having to change any of the existing configurations. In most situations, existing Connect deployments will have already created the storage topics prior to upgrading.
The replication factor defaults to 3, so anyone running Kafka clusters with fewer nodes than 3 will receive an error unless they explicitly set the replication factor for the three internal topics. This is actually desired behavior, since it signals the users that they should be aware they are not using sufficient replication for production use.
The integration tests use a cluster with a single broker, so they were changed to explicitly specify a replication factor of 1 and a single partition.
The `KafkaAdminClientTest` was refactored to extract a utility for setting up a `KafkaAdminClient` with a `MockClient` for unit tests.
Author: Randall Hauch <rhauch@gmail.com>
Reviewers: Ewen Cheslack-Postava <ewen@confluent.io>
Closes#2984 from rhauch/kafka-4667
Author: Damian Guy <damian.guy@gmail.com>
Author: Guozhang Wang <wangguoz@gmail.com>
Author: Apurva Mehta <apurva@confluent.io>
Reviewers: Guozhang Wang, Jason Gustafson, Apurva Mehta, Jun Rao
Closes#2849 from dguy/exactly-once-tc
This PR replaces https://github.com/apache/kafka/pull/2743 (just raising from Confluent repo)
This PR describes the addition of Partition Level Leader Epochs to messages in Kafka as a mechanism for fixing some known issues in the replication protocol. Full details can be found here:
[KIP-101 Reference](https://cwiki.apache.org/confluence/display/KAFKA/KIP-101+-+Alter+Replication+Protocol+to+use+Leader+Epoch+rather+than+High+Watermark+for+Truncation)
*The key elements are*:
- Epochs are stamped on messages as they enter the leader.
- Epochs are tracked in both leader and follower in a new checkpoint file.
- A new API allows followers to retrieve the leader's latest offset for a particular epoch.
- The logic for truncating the log, when a replica becomes a follower, has been moved from Partition into the ReplicaFetcherThread
- When partitions are added to the ReplicaFetcherThread they are added in an initialising state. Initialising partitions request leader epochs and then truncate their logs appropriately.
This test provides a good overview of the workflow `EpochDrivenReplicationProtocolAcceptanceTest.shouldFollowLeaderEpochBasicWorkflow()`
The corrupted log use case is covered by the test
`EpochDrivenReplicationProtocolAcceptanceTest.offsetsShouldNotGoBackwards()`
Remaining work: There is a do list here: https://docs.google.com/document/d/1edmMo70MfHEZH9x38OQfTWsHr7UGTvg-NOxeFhOeRew/edit?usp=sharing
Author: Ben Stopford <benstopford@gmail.com>
Author: Jun Rao <junrao@gmail.com>
Reviewers: Ismael Juma <ismael@juma.me.uk>, Jun Rao <junrao@gmail.com>
Closes#2808 from benstopford/kip-101-v2
Kafka brokers have a config called "offsets.topic.replication.factor" that specify the replication factor for the "__consumer_offsets" topic. The problem is that this config isn't being enforced. If an attempt to create the internal topic is made when there are fewer brokers than "offsets.topic.replication.factor", the topic ends up getting created anyway with the current number of live brokers. The current behavior is pretty surprising when you have clients or tooling running as the cluster is getting setup. Even if your cluster ends up being huge, you'll find out much later that __consumer_offsets was setup with no replication.
The cluster not meeting the "offsets.topic.replication.factor" requirement on the internal topic is another way of saying the cluster isn't fully setup yet.
The right behavior should be for "offsets.topic.replication.factor" to be enforced. Topic creation of the internal topic should fail with GROUP_COORDINATOR_NOT_AVAILABLE until the "offsets.topic.replication.factor" requirement is met. This closely resembles the behavior of regular topic creation when the requested replication factor exceeds the current size of the cluster, as the request fails with error INVALID_REPLICATION_FACTOR.
Author: Onur Karaman <okaraman@linkedin.com>
Reviewers: Jason Gustafson <jason@confluent.io>, Ismael Juma <ismael@juma.me.uk>, Ewen Cheslack-Postava <ewen@confluent.io>
Closes#2177 from onurkaraman/KAFKA-3959
Author: Ismael Juma <ismael@juma.me.uk>
Reviewers: Gwen Shapira <cshapi@gmail.com>, Jason Gustafson <jason@confluent.io>
Closes#2354 from ijuma/kafka-4565-separation-of-internal-and-external-traffic
I spent a bit of time tracking down why files were being deleted before they reached log.retention.hours of age. It turns out that the time and size log retention schemes function independently, and not as the original comment "The minimum age of a log file to be eligible for deletion" might indicate to a new user.
Author: Mark Rose <markrose@markrose.ca>
Reviewers: Guozhang Wang <wangguoz@gmail.com>
Closes#28 from MarkRose/fix_misleading_configuration_file_for_trunk
Author: Guozhang Wang <wangguoz@gmail.com>
Reviewers: Damian Guy <damian.guy@gmail.com>, Matthias J. Sax <matthias@confluent.io>, Michael G. Noll <michael@confluent.io>, Greg Fodor <gfodor@gmail.com>, Ismael Juma <ismael@juma.me.uk>, Ewen Cheslack-Postava <ewen@confluent.io>
Closes#1530 from guozhangwang/K3769-per-thread-metrics
ijuma
As discussed in https://github.com/apache/kafka/pull/1645, this patch removes an extraneous line from several __init__.py files, and a few others as well
Author: Geoff Anderson <geoff@confluent.io>
Reviewers: Ismael Juma <ismael@juma.me.uk>
Closes#1659 from granders/minor-cleanup-init-files
This is a follow up of KAKFA-3421 to update the connect developer guide to include the configuration validation. Also includes a couple of minor fixes.
Author: Liquan Pei <liquanpei@gmail.com>
Reviewers: Jason Gustafson <jason@confluent.io>, Ewen Cheslack-Postava <ewen@confluent.io>
Closes#1366 from Ishiihara/connect-dev-doc
It is not true in practice. Maybe the implied feature is not yet implemented or removed.
These lines can be super misleading.
Please merge.
Thank you.
Author: gaob13 <gaob13@mails.tsinghua.edu.cn>
Reviewers: Ismael Juma, Ewen Cheslack-Postava
Closes#793 from gaob13/trunk
Author: Grant Henke <granthenke@gmail.com>
Reviewers: Ismael Juma <ismael@juma.me.uk>, Jun Rao <junrao@gmail.com>
Closes#686 from granthenke/compaction
Also include some trivial clean-ups in `ProducerConfig`and `BaseProducer`.
Author: Ismael Juma <ismael@juma.me.uk>
Reviewers: Gwen Shapira
Closes#710 from ijuma/use-new-producer-properties-in-config
This adds coordination between DistributedHerders using the generalized consumer
support, allowing automatic balancing of connectors and tasks across workers. A
few pieces that require interaction between workers (resolving config
inconsistencies, forwarding of configuration changes to the leader worker) are
incomplete because they require REST API support to implement properly.
Author: Ewen Cheslack-Postava <me@ewencp.org>
Reviewers: Jason Gustafson, Gwen Shapira
Closes#321 from ewencp/kafka-2371-distributed-herder
This also adds some other needed infrastructure for distributed Copycat, most
importantly the DistributedHerder, and refactors some code for handling
Kafka-backed logs into KafkaBasedLog since this is shared betweeen offset and
config storage.
Author: Ewen Cheslack-Postava <me@ewencp.org>
Reviewers: Gwen Shapira, James Cheng
Closes#241 from ewencp/kafka-2372-copycat-distributed-config
Author: Parth Brahmbhatt <brahmbhatt.parth@gmail.com>
Reviewers: Ismael Juma <ismael@juma.me.uk>, Jun Rao <junrao@gmail.com>
Closes#195 from Parth-Brahmbhatt/KAFKA-2211
The Converter class now translates directly between byte[] and Copycat's data
API instead of requiring an intermediate runtime type like Avro's GenericRecord
or Jackson's JsonNode.
Author: Ewen Cheslack-Postava <me@ewencp.org>
Reviewers: Gwen Shapira
Closes#172 from ewencp/kafka-2475-unified-serializer-converter and squashes the following commits:
566c52f [Ewen Cheslack-Postava] Checkstyle fixes
320d0df [Ewen Cheslack-Postava] Restrict offset format.
85797e7 [Ewen Cheslack-Postava] Add StringConverter for using Copycat with raw strings.
698d65c [Ewen Cheslack-Postava] Move and update outdated comment about handing of types for BYTES type in Copycat.
4bed051 [Ewen Cheslack-Postava] KAFKA-2475: Make Copycat only have a Converter class instead of Serializer, Deserializer, and Converter.
This is an initial patch implementing the basics of Copycat for KIP-26.
The intent here is to start a review of the key pieces of the core API and get a reasonably functional, baseline, non-distributed implementation of Copycat in place to get things rolling. The current patch has a number of known issues that need to be addressed before a final version:
* Some build-related issues. Specifically, requires some locally-installed dependencies (see below), ignores checkstyle for the runtime data library because it's lifted from Avro currently and likely won't last in its current form, and some Gradle task dependencies aren't quite right because I haven't gotten rid of the dependency on `core` (which should now be an easy patch since new consumer groups are in a much better state).
* This patch currently depends on some Confluent trunk code because I prototyped with our Avro serializers w/ schema-registry support. We need to figure out what we want to provide as an example built-in set of serializers. Unlike core Kafka where we could ignore the issue, providing only ByteArray or String serializers, this is pretty central to how Copycat works.
* This patch uses a hacked up version of Avro as its runtime data format. Not sure if we want to go through the entire API discussion just to get some basic code committed, so I filed KAFKA-2367 to handle that separately. The core connector APIs and the runtime data APIs are entirely orthogonal.
* This patch needs some updates to get aligned with recent new consumer changes (specifically, I'm aware of the ConcurrentModificationException issue on exit). More generally, the new consumer is in flux but Copycat depends on it, so there are likely to be some negative interactions.
* The layout feels a bit awkward to me right now because I ported it from a Maven layout. We don't have nearly the same level of granularity in Kafka currently (core and clients, plus the mostly ignored examples, log4j-appender, and a couple of contribs). We might want to reorganize, although keeping data+api separate from runtime and connector plugins is useful for minimizing dependencies.
* There are a variety of other things (e.g., I'm not happy with the exception hierarchy/how they are currently handled, TopicPartition doesn't really need to be duplicated unless we want Copycat entirely isolated from the Kafka APIs, etc), but I expect those we'll cover in the review.
Before commenting on the patch, it's probably worth reviewing https://issues.apache.org/jira/browse/KAFKA-2365 and https://issues.apache.org/jira/browse/KAFKA-2366 to get an idea of what I had in mind for a) what we ultimately want with all the Copycat patches and b) what we aim to cover in this initial patch. My hope is that we can use a WIP patch (after the current obvious deficiencies are addressed) while recognizing that we want to make iterative progress with a bunch of subsequent PRs.
Author: Ewen Cheslack-Postava <me@ewencp.org>
Reviewers: Ismael Juma, Gwen Shapira
Closes#99 from ewencp/copycat and squashes the following commits:
a3a47a6 [Ewen Cheslack-Postava] Simplify Copycat exceptions, make them a subclass of KafkaException.
8c108b0 [Ewen Cheslack-Postava] Rename Coordinator to Herder to avoid confusion with the consumer coordinator.
7bf8075 [Ewen Cheslack-Postava] Make Copycat CLI speific to standalone mode, clean up some config and get rid of config storage in standalone mode.
656a003 [Ewen Cheslack-Postava] Clarify and expand the explanation of the Copycat Coordinator interface.
c0e5fdc [Ewen Cheslack-Postava] Merge remote-tracking branch 'origin/trunk' into copycat
0fa7a36 [Ewen Cheslack-Postava] Mark Copycat classes as unstable and reduce visibility of some classes where possible.
d55d31e [Ewen Cheslack-Postava] Reorganize Copycat code to put it all under one top-level directory.
b29cb2c [Ewen Cheslack-Postava] Merge remote-tracking branch 'origin/trunk' into copycat
d713a21 [Ewen Cheslack-Postava] Address Gwen's review comments.
6787a85 [Ewen Cheslack-Postava] Make Converter generic to match serializers since some serialization formats do not require a base class of Object; update many other classes to have generic key and value class type parameters to match this change.
b194c73 [Ewen Cheslack-Postava] Split Copycat converter option into two options for key and value.
0b5a1a0 [Ewen Cheslack-Postava] Normalize naming to use partition for both source and Kafka, adjusting naming in CopycatRecord classes to clearly differentiate.
e345142 [Ewen Cheslack-Postava] Remove Copycat reflection utils, use existing Utils and ConfigDef functionality from clients package.
be5c387 [Ewen Cheslack-Postava] Minor cleanup
122423e [Ewen Cheslack-Postava] Style cleanup
6ba87de [Ewen Cheslack-Postava] Remove most of the Avro-based mock runtime data API, only preserving enough schema functionality to support basic primitive types for an initial patch.
4674d13 [Ewen Cheslack-Postava] Address review comments, clean up some code styling.
25b5739 [Ewen Cheslack-Postava] Fix sink task offset commit concurrency issue by moving it to the worker thread and waking up the consumer to ensure it exits promptly.
0aefe21 [Ewen Cheslack-Postava] Add log4j settings for Copycat.
220e42d [Ewen Cheslack-Postava] Replace Avro serializer with JSON serializer.
1243a7c [Ewen Cheslack-Postava] Merge remote-tracking branch 'origin/trunk' into copycat
5a618c6 [Ewen Cheslack-Postava] Remove offset serializers, instead reusing the existing serializers and removing schema projection support.
e849e10 [Ewen Cheslack-Postava] Remove duplicated TopicPartition implementation.
dec1379 [Ewen Cheslack-Postava] Switch to using new consumer coordinator instead of manually assigning partitions. Remove dependency of copycat-runtime on core.
4a9b4f3 [Ewen Cheslack-Postava] Add some helpful Copycat-specific build and test targets that cover all Copycat packages.
31cd1ca [Ewen Cheslack-Postava] Add CLI tools for Copycat.
e14942c [Ewen Cheslack-Postava] Add Copycat file connector.
0233456 [Ewen Cheslack-Postava] Add copycat-avro and copycat-runtime
11981d2 [Ewen Cheslack-Postava] Add copycat-data and copycat-api
Randall Hauch
Konstantine Karantasis
Edoardo Comar
Damian Guy
Ben Stopford
Akhilesh Naidu
Onur Karaman
Ismael Juma
Mark Rose
Guozhang Wang
Geoff Anderson
Liquan Pei
gaob13
Ashish Singh
Jason Gustafson
Grant Henke
Gwen Shapira
Ewen Cheslack-Postava
Parth Brahmbhatt
Manikumar Reddy
Raman Gupta
James Oliver
Anton Karamanov
Sriharsha Chintalapani
Joe Stein
Jay Kreps
Neha Narkhede
Joel Koshy
Jun Rao
Roger Hoover
Swapnil Ghike