On Wed, Sep 16, 2009 at 6:24 PM, Jon Graham <[email protected]> wrote:
> What is the recommended method for backing up or check pointing HBase? > > + We just committed mapreduce jobs to export and then import designated table (See Import and Export under the mapreduce package in trunk and branch). + You could quiesce the table (or better disable it, or make it read-only), flush it, then copy off the content at /HBASE.ROOTDIR/TABLENAME. To restore, we'd need to put together a script for you. It would read the necessary metadata from the copy and populate the hbase catalog table with the copies list of regions (we write under each region a '.regioninfo' file to facilitate this kinda thing). See https://issues.apache.org/jira/browse/HBASE-50 for more on ideas like this and for thoughts on a formal checkpoint. > We regularly backup the Hadoop NameNode check points and definitely want to > avoid any > database corruption or reconstruction issues. > > Is there an HBase tool similar to Hadoop's fsck? > There are bits and pieces -- see HFile.main -- and some admin tools in shell -- type 'tools' -- but we need to pull them all together better (See HBASE-7). St.Ack > > Thanks, > Jon >
