Apache9 commented on pull request #2237: URL: https://github.com/apache/hbase/pull/2237#issuecomment-673896406
The inconsistency is between the data on filesystem and zookeeper. In general, after the meta table has been initialized, it should have a directory on the file system, and also a znode to indicate its state on zookeeper. The problem here is that, in your scenario, the data on zookeeper are all gone. This is what I called 'inconsistency'. And I've also mentioned several times, it works for several versions does not mean it will work for all the versions. They are just different implementation choices, and you can not force the developpers to not use the data on zookeeper to determine what to do next. So I do not think it worth to ask on the dev list about 'whether' the InitMetaProcedure should delete meta without checking 'partial' or not. The problem here is you delete the state znode for meta region on zookeeper, not a problem for InitMetaProcedure. Either you add it back through HBCK2, or you should have a overall design on how we could fix the problem, without using HBCK2. ---------------------------------------------------------------- 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]
