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: [One of " Under Discussion", "Accepted", "Rejected"]
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]
JIRA:
Jira | ||||||
---|---|---|---|---|---|---|
|
Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).
Motivation
Describe the problems you are trying to solve.
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
MirrorMaker 1.0 currently inherits the default value for `auto.offset.reset`, which is `latest`.
While for most consumers this is a sensible default, MirrorMakers are specifically designed for replication, so they should default to replicating topics from the beginning.
A specific scenario where the current behavior causes date "loss" is when a MirrorMaker is subscribed to a regex pattern. If auto-topic creation is enabled on the cluster, and you start producing to a non-existent topic that matches the regex, then there will be a period of time where the producer is producing before the new topic's partitions have been picked up by the MirrorMaker. Those messages will never be consumed by the MirrorMaker because it will start from latest, ignoring those just-produced messages.
In fact, the new MirrorMaker 2.0 sets exactly this config: https://github.com/apache/kafka/blob/d63eaaaa0181bb7b9b4f5ed088abc00d7b32aeb0/connect/mirror/src/main/java/org/apache/kafka/connect/mirror/MirrorConnectorConfig.java#L233
So this change will simply bring the old MirrorMaker 1.0 into compliance with the behavior of the new MirrorMaker 2.0 which already behaves this way.
Proposed Changes
This would add a MirrorMaker 1.0 default consumer property of `auto.offset.reset==earliest`. Users can still override this in the MirrorMaker consumer config fileDescribe the new thing you want to do in appropriate detail. This may be fairly extensive and have large subsections of its own. Or it may be a few sentences. Use judgement based on the scope of the change.
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?
Rejected Alternatives
This will be a silent breaking change since it flips the behavior around.
Existing mirrormakers will be unaffected for any topics they are currently consuming since they already have a saved offset.
Mirrormakers that start consuming topics for which they don't have a saved offset will start replicating the partitions from the beginning, rather than from the partition's current highwater mark. If the mirrormaker starts consuming a very large partition/topic, it will replicate far more data than previously. This has relatively low probability since most topics that get replicated are newly-created, in which case starting from the earliest simply prevents skipping the first few seconds/minutes of data written to the topic.
Since MirrorMaker 2.0 already behaves this way, this change will make future migrations from MM1 to MM2 easier for folks since the behavior will stop changing between them.
Rejected Alternatives
Leaving it as-is. As noted in the description, the existing state of affairs produces data gaps for anyone replicating topics using a regex patternIf there are alternative ways of accomplishing the same thing, what were they? The purpose of this section is to motivate why the design is the way it is and not some other way.