Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Table of Contents

This page is meant as a template for writing a KIP. To create a KIP choose Tools->Copy on this page and modify with your content and replace the heading with the next KIP number and a description of your issue. Replace anything in italics with your own description.

Status

Current state: DraftAccepted

Discussion thread: here [Change the link from the KIP proposal email archive to your own email thread]

JIRA: here [Change the link from KAFKA-1 to your own ticket]

Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).

Motivation

One of MM2's main feature features is syncing consumer group offset between clusters. This feature relays relies on two separate connectors- `MirrorSourceConnector`

  1. MirrorSourceConnector : this connector stores the translation of

...

  1. the new record offsets on target cluster in

...

  1. offset_

...

  1. syncs  internal topic after mirroring the record itself. This translation is done at

...

  1. commit  method since 3.5 where

...

  1. commit()

...

  1.   API stores the offsets in the source system, up to the offsets that have been returned by

...

  1. poll()

...

  1. . The implementation of this API should block until the commit is complete. See

...

  1. KAFKA-14610

...

  1. for more details.

...

  1. MirrorCheckpointConnector : this connector uses the new record offsets from

...

  1. offset_

...

  1. syncs  and update

...

  1. checkpoints  and

...

  1. __consumer_

...

  1. offsets  topics. This can be controlled by

...

  1. emit.checkpoints.enabled  and sync.group.offsets.

...

  1. enabled 

For use-cases where customers only run MM2 for replicating between clusters without the need for moving the consumer group offsets by running only MirrorSourceConnector still left with part#1 .  Part#1 (blocking until commit) of this feature which add still an issue as it adds cost to the progress of their replication. The only workaround reducing the latency cost is by increasing offset.lag.max  to INT_MAX  which will result that the offset will never going to be queued for offset sync (Check `PartitionState::update` for details on how MM2 decided if it should emit an offset sync or not) however this doesn't eliminate the cost of creating an internal topic. 

This KIP propose proposes that we add a new config for `MirrorSourceConnector` MirrorSourceConnector  to control enabling the offsets translation of offsets between clusters for replicated topics.

Public Interfaces

Briefly list any new interfaces that will be introduced as part of this proposal or any existing interfaces that will be removed or changed. The purpose of this section is to concisely call out the public contract that will come along with this feature.

A public interface is any change to the following:

  • Binary log format

  • The network protocol and api behavior

  • Any class in the public packages under clientsConfiguration, especially client configuration

    • org/apache/kafka/common/serialization

    • org/apache/kafka/common

    • org/apache/kafka/common/errors

    • org/apache/kafka/clients/producer

    • org/apache/kafka/clients/consumer (eventually, once stable)

  • Monitoring

  • Command line tools and arguments

  • Anything else that will likely break existing users in some way when they upgrade

Proposed Changes

...

PropertiesDefaultDescription
emit.offset-syncs.enabled true

Whether to store the new offset of the replicated records in offset-syncs topic or not. MirrorCheckpointConnector will fail to start if emit.checkpoints.enabled and/or sync.group.offsets.enabled are enabled while emit.offset-syncs.enabled  are disabled.

Proposed Changes

  • Similar to emit.checkpoints.enabled   and emit.heartbeat.enabled   this KIP will add emit.offset-syncs.enabled . The config default value will be true to keep the current implementation.

  • offset-syncs topic will not be created if emit.offset-syncs.enabled  set to false 

  • logic to publish offset syncs in commit()  will not be executed if emit.offset-syncs.enabled  set to false 

  • MirrorCheckpointConnector  validation will fail at start if emit.offset-syncs.enabled  disabled while emit.checkpoints.enabled  and/or sync.group.offsets.enabled  are enabled. 

Compatibility, Deprecation, and Migration Plan

  • What impact (if any) will there be on existing users?
  • If we are changing behavior how will we phase out the older behavior?
  • If we need special migration tools, describe them here.
  • When will we remove the existing behavior?

Test Plan

Describe in few sentences how the KIP will be tested. We are mostly interested in system tests (since unit-tests are specific to implementation details). How will we know that the implementation works as expected? How will we know nothing broke?

Rejected Alternatives

...

emit.offset-syncs.enabled  is set to true by default so any existing use-cases or use-cases that run this feature will do nothing. Only use-case that wish to fully disable this feature will need to set emit.offset-syncs.enabled to false.

Rejected Alternatives

  • Reuse emit.checkpoints.enabled   and sync.group.offsets.enabled  and assume that if both are disabled then this feature needs to be fully disabled. I didn't want to rely on MirrorCheckpointConnector  configs to configure MirrorSourceConnector . And this might impact customers who enabled  MirrorSourceConnector  with intention to run MirrorCheckpointConnector  later in the future