hi Jian

If the issue is not critical, it would be better to have more discussion before 
vote :)

Best,
Chia-Ping

> jian fu <[email protected]> 於 2025年11月7日 清晨7:42 寫道:
> 
> Hi all,
> 
> I’d like to initiate the vote for KIP-1235 right away as it’s a minor
> change with minimal risk.
> 
> Summary:
> 
> Currently, __remote_log_metadata relies on the broker default for min.isr
> (typically 1).
> With min.isr=1 and factor=3, acknowledged writes can be lost if the leader
> crashes before replication completes, making remote data permanently
> unreachable.
> What's more, the similar topic: __transaction_state sets min.isr=2 via
> transaction.state.log.min.isr.We should provide the same durability for
> __remote_log_metadata.
> 
> Proposed Change:
> Add one new broker config: remote.log.metadata.topic.min.isr (default: 2)
> When creating __remote_log_metadata, TopicBasedRemoteLogMetadataManager
> will set min.insync.replicas=2 at the topic level.
> 
> KIP WIKI:
> https://cwiki.apache.org/confluence/display/KAFKA/KIP-1235%3A+Correct+the+default+min.insync.replicas+to+2+for+the+__remote_log_metadata+topic
> Discussion Thread:
> https://lists.apache.org/thread/016mjw3rrcqnt8475bxr3b3npm0t32xl
> PR: https://github.com/apache/kafka/pull/20811
> 
> Looking forward to your feedback!
> 
> Thanks!
> 
> Regards
> Jian

Reply via email to