Just file issues if you find anything wrong in the future. I think if we have the same goal, we could finally work it out :)
Thanks~ Nick Dimiduk <ndimi...@apache.org> 于2021年9月14日周二 上午1:12写道： > Sadly, I’ve not had time to review this effort. I do hope someone has given > it the attention it deserves. I look forward to these class hierarchies > finally being disentangled! > > Thanks, > Nick > > On Sun, Sep 12, 2021 at 07:35 张铎(Duo Zhang) <palomino...@gmail.com> wrote: > > > Just a reminder, will merge this PR tomorrow if no more concerns. > > > > Thanks. Regards > > > > 张铎(Duo Zhang) <palomino...@gmail.com> 于2021年9月3日周五 下午10:19写道： > > > > > The related issue is HBASE-15549, where we decided to not let HMaster > > > carry regions. > > > > > > After some prior works, such as modifying balancer, refactoring > > connection > > > registry, replementing maintenance mode related code, etc, now we > arrive > > > the last step, where we could finally decouple HMaster and > HRegionServer, > > > which means now HMaster does not need to extends HRegionServer any > more. > > > > > > This could greatly reduce the complexity of the code in both HMaster > and > > > HRegionServer. The current code is very very flaky as it is very easy > to > > > introduce cyclic dependencies and hang the start up of HMaster, as when > > > starting a HRegionServer, we need to register to HMaster first but > > HMaster > > > itself is a HRegionServer... > > > > > > The last PR is > > > > > > https://github.com/apache/hbase/pull/3612 > > > > > > It is kinda big so I posted this email here to attract more reviewers. > > > > > > Feel free to reply here directly or on the PR about your thoughts. > > > > > > Thanks. Regards > > > > > >
