HBase runs off the road if you enable/disable/enable/disable the table and then
delete
--------------------------------------------------------------------------------------
Key: HBASE-3077
URL: https://issues.apache.org/jira/browse/HBASE-3077
Project: HBase
Issue Type: Bug
Reporter: stack
HBASE-3063 has a bit of detail. The new master is better around disable,
delete but still not good enough. This issue is about digging in a bit. While
disable and enable are not just flags up in zk, setting either is followed by
the actual open/close of regions. We need to add support to the master where
it can follow swift enabling/disabling. HBASE-3063 add some lag to the delete
so we'll wait on region close before we go ahead and remove the region but we
should be able to do better here too... e.g. make use of Nicolas's new
interrupt compactions so delete doesn't ahve to wait on ongoing compactions...
and we could add an abort of specific regions only so we could just crash them
closed rather than wait on flushes if they are going to be just deleted anyways.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
