Status

Current state: Under Discussion

Discussion thread: here [TBD]

JIRA: KAFKA-6989 

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

Motivation

Currently, in Kafka Streams, a single thread is used to process tasks which could result in a performance bottleneck. With the current model, under the condition of a fatal error, a new retry topic will be created in which failed records will be sent. Incoming records could not be processed until the failed records in the retry topic are fully reprocessed to guarantee ordering. This could result in a lag because we are essentially backtracking to make sure all data has been sent. New data that would have been processed if a failure had not occurred will have to wait while the thread caught up.

Public Interfaces

We have a couple of choices available to us as mentioned in the ticket discussion (KAFKA-6989). Recall that in asynchronous methods, we do not need inputs from a previous function to start it. All we need is the signal. However, when offsets have been committed in an asynchronized manner, we will need to define way to determine the behavior that occurs after the conclusion of the process. It would also be wise to remember that asynchronous processing does not necessarily require an extra thread.  When offsets are committed or received, we should consider the following:

1. Ordering: As noted in the JIRA chat, Samza has judged it to be impossible for the records to be returned in its original sequence. A sort of pseudo-ordering could be guaranteed if we sort the records that we return by time of arrival and offset number, but that might not matter considering that the record's offsets are not consecutive. The other option is to block while only sending the records from the "retry" topic until they have all been processed. See example below.
2. Exactly-Once: In exactly-once semantics, 

Proposed Changes

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?

Rejected Alternatives

If 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.