Re: [DISCUSS] KIP-1176: Tiered Storage for Active Log Segment

er partition and > what's the failover time? > > JR8. Is every broker consuming __remote_wal_log_metadata? Since the volume > of the metadata is much higher than tier storage, what's the additional > network cost for metadata propagation? > > Jun > > > On Mon,

Re: [DISCUSS] KIP-1176: Tiered Storage for Active Log Segment

tentially a more critical > > problem here since it's dealing with data in the active segment. > > > > JR6. "S3E1Z and EBS are the choice in AWS world" > > How does EBS work? EBS can't read from a different availability zone, > > right? > > >

Re: [DISCUSS] The Path Forward for Saving Cross-AZ Costs KIPs

Hi Luke, Thanks for creating this discussion on these KIPs. I'm collaborating with Henry on KIP-1176. We have deployed the Kafka fork to one of our Kafka clusters. We collected performance data (full results here

Edit Permissions for Existing KIP Confluence Page

Hi, I'm collaborating on KIP-1176 . I'm a colleague of the original page creator: Henry Cai. I talked to him and he can't grant edit permission to me. Are you all able to grant edit permissions on th

Re: [DISCUSS] KIP-1176: Tiered Storage for Active Log Segment

Hi Jun, Thanks for the feedback! Sorry for the delayed response, we didn't see this until today. We will also add relevant parts from this response to the KIP-1176 confluence page. JR1: This is somewhat similar to the current situation where we put 3 brokers in 3 distinct AZs. When AZ1 is down th

Re: [DISCUSS] KIP-1176: Tiered Storage for Active Log Segment

Hi Christo, Thanks for your feedback! Responding as a co-author to some of these items: For acks=-1, it's true with AZ failure S3E1Z data could be lost. In this case, we can use option 1 "The leader broker will wait until all in-sync follower replicas have reached the given offset before the ackn