Apache9 commented on pull request #2584: URL: https://github.com/apache/hbase/pull/2584#issuecomment-722153134
> > Landing this on master is proposed by me as this PR is not related to the server side changes. It can be used in our current code base, without the changes in HBASE-18070. And what's more, the client side code are different between master and branch-2, as on master, we rebuild the sync client on top of async client which makes it much easier to implement this issue but on branch-2, you need to deal with sync client separately. So I suggest we land this on master, and then start backporting to branch-2 ASAP. > > As I understand, HBASE-18070 is branched from master. As we are merging HBASE-18070 back to master, it would be better to merge them as a whole. Unitests are different w/o meta replication source changes in HBASE-18070 as there is no realtime replication of meta wal edits. I simulated that by "flush" and "refresh" hfiles for meta. > Would do you think? I‘m fine with committing to HBASE-18070 but it will delay the whole process I suppose. If you commit this to master, while backporting to branch-2, you just need to backport this commit only. If you commit to HBASE-18070, while backporting you need to deal with all the commits. And you can change the UT on the feature branch when meta replication is available. In general, I just want to help you guys land the feature faster. I do not get the point why you do not want to commit to master, usually developpers will do the opposite... Anyway, I've approved the PR. You are free to commit to master or feature branch as you like. Thanks. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
