[ https://issues.apache.org/jira/browse/HBASE-20666?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16587896#comment-16587896 ]
Ted Yu commented on HBASE-20666: -------------------------------- For #1, I see that you have an assertion in finally block which is equivalent to what I suggested. Basically you can put Assert.fail() call following table creation since you know that table creation would throw exception. > Unsuccessful table creation leaves entry in rsgroup meta table > -------------------------------------------------------------- > > Key: HBASE-20666 > URL: https://issues.apache.org/jira/browse/HBASE-20666 > Project: HBase > Issue Type: Bug > Reporter: Biju Nair > Assignee: Xu Cang > Priority: Minor > Attachments: HBASE-20666.master.001.patch, > HBASE-20666.master.002.patch > > > If a table creation fails in a cluster enabled with {{rsgroup}} feature, the > table is still listed as part of {{default}} rsgroup. > To recreate the scenario: > - Create a namespace (NS) with number of region limit > - Create table in the NS which satisfies the region limit by pre-splitting > - Create a new table in the NS which will fail > - {{list_rsgroup}} will show the table being part of {{default}} rsgroup and > data can be found in {{hbase:rsgroup}} table > Would be good to revert the entry when the table creation fails or a script > to clean up the metadata. -- This message was sent by Atlassian JIRA (v7.6.3#76005)