Apache9 commented on pull request #2237: URL: https://github.com/apache/hbase/pull/2237#issuecomment-673836007
I think we'd better start from a high level design, on how we could start a cluster with no data on zookeeper. My proposal is always, introducing a new option in HBCK2 to reconstruct the data on zookeeper, but you actually start the cluster. If you guys think it is possible to do this in normal code path, please give an overall design. Open a google doc? So others could comments on it? For deleting meta, I agree that we should not delete it if it is not 'partial'. But as I said above many times, if there is no inconsitency, scheduling an InitMetaProcedure implies that we do not have a meta table yet, so in InitMetaProcedure it is always safe to delete the meta directory. So even if we add a check in InitMetaProcedure to determine whether the meta table is 'real' partial, the result should be an ERROR log to tell users something crtical happens, please fix the inconsistency before moving forward. I do not think this is what you guys want here? As you want to start the HMaster and recover from the inconsistency? ---------------------------------------------------------------- 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: us...@infra.apache.org
