HBASE-3008 Memstore.updateColumnValue passes wrong flag to heapSizeChange incrementColumnValue cannot be used in this released due to a bug caused by a portion of a patch that was committed in HBASE-3008. It's causing the memstores to grow in negative size, disabling flushes and losing data when the region closes. HBASE-2998 rolling-restart.sh shouldn't rely on zoo.cfg If HBase wasn't configured with zoo.cfg (meaning that the ZK configurations are set in hbase-site.xml), the script will likely kill the whole cluster instead of rolling restart it.