David Jacot
David Jacot
@chia7712 Done. Sorry for that.
@chia7712 Right. I just removed it.
There are a few failed tests that I need to fix…
Hi @FrankYang0529. Thanks for working on this. I have a few high level comments/questions to start with. 1. I am not convinced that using `hashCode()` is the right approach. Are...
> Agree, hashCode() is not good enough for it. We should use something like md5. Also, the hash function should generate same result for same data with different order, because...
@FrankYang0529 @chia7712 We may need a KIP for this one. Let's first agree on the ideal solution and we can do a small KIP if needed.
@FrankYang0529 When you update the PR, I wonder whether we could split it into smaller chunks. That would ease the reviews. Would it be possible?
If it is not enough, you can try to reduce it further (e.g. 1MB) and you can also reduce the number of partitions with [offsets.topic.num.partitions](https://kafka.apache.org/documentation/#brokerconfigs_offsets.topic.num.partitions). It is 50 by default....
@FrankYang0529 Thanks for the patch! Have you already tried to run the micro-benchmarks? The performance of the assignors is critical so we must take a performance driven approach.