05-14-2017 09:14 PM
when spark can use the metadata (KUDU table) which is create by impala?
now we using spark to access the kudu table create by the impala, can't direct using the hive metadata(create by impala),
beause the storage handle set by impala is not really exist, it's seems only used by impala to check it's a kudu table or not.
so now we access the kudu table direct by refer to kudu service, but this method is bypass the security guard which normally protected by sentry!