Hi Matteo, Thanks a lot for your quick and detailed explanation.
-OC On Tue, Nov 19, 2013 at 10:43 PM, Matteo Bertozzi <theo.berto...@gmail.com>wrote: > > 1.Is there a way to take incremental snapshots on HBase tables? > > > > Due to the nature of hbase, when hfiles gets compacted your table will not > share files with the snapshot causing the next snapshot to don't share > hfiles (or share a few) with the previous snapshot. The other approach will > be using logs to have an incremental snapshots but logs contains different > tables, so you may end up with more space used, and for sure more time to > restore since you must replay the logs. > > > > 2.Can we export snapshots to outside of HDFS(e.g backup server)? > > > > the ExportSnapshot tool takes a destination path as argument, so as long as > the destination filesystem is supported by the FileSystem class you can. > (e.g. you can export with -copy-to=file:///mnt/backup-mount-point/hbase) > > > > 3.Is it possible to take a snapshot for all tables in one command? > > > > No, but you can do a simple script that does a for each table in > admin.listTable(), admin.snapshot(table, snapshotName) > > > > 4.Does HBase support to take automatic snapshots based on configured > > timeout? > > > > No, hbase doesn't do job scheduling. > you may use a cron job to execute a shell command, "echo snapshot > 'tableName', 'snapshotName' | hbase shell" > > > > > > Please also let us know if there is any better way to take HBase tables > > backup. > > > > There are different solutions that you can use: Replication, CopyTable, > HDFS files copy + CopyTable and maybe others it depends on what you're > looking for. > There is an article here that explains how to use CopyTable as backup > tool: > > http://blog.cloudera.com/blog/2012/06/online-hbase-backups-with-copytable-2/ >
