Status

Current state:  Under discussionAccepted

Vote thread: https://lists.apache.org/thread/vvw45tcdmb23d8gpgb3x4fklx3tp7hpm

Discussion thread: TODO
JIRA: TODO: https://lists.apache.org/thread/w4ml9ffkj1j31j8kjpbywq9jsw5ck5sr

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

Currently, there exists a kafka.server:type=MetadataLoader,name=CurrentMetadataVersion  metric which tells the operator the metadata version feature level for every node in a cluster. However, now that there several features with support beyond the metadata.version  feature, adding a metric to display those features' levels would be helpful for monitoring upgrade/downgrade scenarios for clusters using those features. Additionally, adding metrics for the minimum and maximum supported feature levels of each node would be helpful for the operator to determine if a feature upgrade/downgrade is safe to perform.

Public Interfaces

Monitoring

Add metrics for each production feature (i.e. metadata.version , kraft.version , transaction.version , group.version , eligible.leader.replicas.version , share.version, and streams.version ).

This can be accomplished via generic "feature level" metrics that are tagged by feature name. Using generic metrics allows for new production features to automatically expose their feature level and supported feature levels going forward. 

The FinalizedLevel  metric will report the finalized feature level for each production feature if it exists. If a feature does not have a finalized level, it will not have an associated FinalizedLevel metric. Similarly, if the finalized level of the metric is "removed" by setting it to 0, the associated FinalizedLevel metric will also be removed. This does not apply to metadata.version , whose minimum version is 7. This metric should still reside in the MetadataLoader metric group because its value is derived from the metadata log's feature records.

minimum-supported-level  and maximum-supported-level  are metrics whose values are dependent only on the software version being run, so it seems appropriate to define a new metric type to house these sorts of metrics going forward (release version, commit hash, etc.)Describe 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?

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

We will deprecate the kafka.server:type=MetadataLoader,name=CurrentMetadataVersion metric since this per-feature tagged metric also exposes this value, and the previous metric can be removed starting in 5.0.

Test Plan

We will add junit tests to verify the new metrics.

Alternatives

Running kafka-feature describe  on every node is one way to track feature level upgrades/downgrades, but it is not straightforward to monitor since it is not a metric.

This KIP discusses the above approach: KIP-1160: Enable returning supported features from a specific brokerIf 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.