[ https://issues.apache.org/jira/browse/HBASE-17825?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16356372#comment-16356372 ]
Ted Yu commented on HBASE-17825: -------------------------------- {code} - new Path(HBackupFileSystem.getTableBackupDataDir(backupRoot, mergedBackupId, tableName)); + new Path(HBackupFileSystem.getTableBackupDir(backupRoot, mergedBackupId, tableName)); {code} Why change directory ? {code} + boolean result = fs.delete(p, false); + if (!result) { + return false; {code} When false is returned, some files would be deleted while some are left. Is this desirable ? {code} + String[] tableNameParts = tableName.split(":"); + Path tableRelPath = new Path(tableName.split(":")[0]); {code} You can use tableNameParts instead of calling split() multiple times. {code} + byte[] outKey = + Bytes.add(table.getName(), Bytes.toBytes(tableSeparator), CellUtil.cloneRow(cell)); {code} Is the above a compatible change ? {code} + List<TableName> tableNames = getTableNameList(tables); {code} After you obtain the List, you iterate thru tableNames. Looks like creation of the List can be avoided. > Backup: further optimizations > ----------------------------- > > Key: HBASE-17825 > URL: https://issues.apache.org/jira/browse/HBASE-17825 > Project: HBase > Issue Type: Improvement > Reporter: Vladimir Rodionov > Assignee: Vladimir Rodionov > Priority: Critical > Labels: backup > Fix For: 3.0.0 > > Attachments: HBASE-17825-v1.patch > > > Some phases of backup and restore can be optimized: > # WALPlayer support for multiple tables > # Run DistCp once per all tables during backup/restore > The eventual goal: > # 2 M/R jobs per backup/restore -- This message was sent by Atlassian JIRA (v7.6.3#76005)