Ethan Rose

Yes, I'm working on a re-review to get this through.

@captainzmc we still need some more work to get this PR ready. I'm not sure it needs to be a release blocker. The bug is if there is an IO...

@smengcl @prashantpogde PTAL when you get a chance for feedback on this approach. This is pending testing with live Ranger as well. If we want to change the default frequency...

Hi @symious, could you please explain how the pending delete block count diverges from the correct value? It seems like we should identify the cause of the divergence before deciding...

> Moving the DeletedBlocksTransaction to common hdds.proto is an ideal solution, but it will lead to the compatibility problem, so I adds the import of ScmServerDatanodeHeartbeatProtocol.proto I don't think we...

Thanks for the proto update @Xushaohong. I probably won't have time to review the rest of this PR but that part looks good. Just a question: I see the new...

> I think the placement in ScmServerDatanodeHeartbeatProtocol.proto is a historical problem. Originally it is only considered to be used in HB, not in client queries. Such common proto shall be...

I see some previous discussion here but I just want to clarify before I +1. This is saying we cannot go from Ratis 1 to Ratis 3 SCM due to...

Ok we can continue with the proposal to remove the deleted containers from RocksDB. Does this description line up with what is being proposed? - When replication manager finds an...

Hi @JacksonYao287 thanks for continuing the work on this. I do still have some concerns with this proposal though. > first, in this patch, DELETED container is removed from scm...