maobaolong commented on pull request #1083: URL: https://github.com/apache/hadoop-ozone/pull/1083#issuecomment-645714602
@avijayanhwx Thanks for your question, it let me think more. > Removing useless column family is a schema change, and schema changes across versions are generally handled by Upgrade handlers in systems. In fact, the column family S3Table is a good example, the schema is removed in OM, but it still exist in the rocksdb, so it doesn't be removed by upgrade handlers, i have to remote it manually though rocksdb/ldb, i thinks this PR is nothing but only the substitute of rocksdb/ldb. > these types of destructive operations can be done possibly by naive users. I just did the drop column family operation for S3Table last week on our Ozone cluster. > But at least in that case, Ozone did not provide a direct way for the user to mess up its own DB. In our Ozone cluster, nobody can login with the admin user to the OM node, even someone can login with the developer user, they can only read the log and rocksdb file, no write permission at all, so they cannot do the drop table, create table, remove/move rocksdb. In this sense, i think it is safety. > Also, in the case of OM HA, doesn't the user have to do it in every OM individually? Yeah, this PR is the only substitute of rocksdb/ldb, and it operate the rocksdb directly, so the operator need to execute this command individually. ---------------------------------------------------------------- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
