Which system went down? HBase or HDFS? or both? Once the dfs.replication set to something larger than one, then the data replication is already being processed by HDFS. So in my best knowledge, hbase store is missing the data block, but it might still exist in the hdfs as a hfile. you could try hbck to resynchronize the data btw hdfs and hbase. wiping out all data is not always the best solution, especially for production clusters.
HBase replication is somewhat different from hdfs replication in that hbase tries "master push" all writes to other slave clusters. In order to use hbase replication, you will need to setup multiple hbase clusters, and explicitly add those clusters as peers and start replication from hbase cmd-line. Hope this helps, -- *Benjamin Kim** benkimkimben at gmail* On Mon, May 28, 2012 at 7:21 PM, kim young ill <[email protected]>wrote: > hi there, > is it possible to configure hbase to automatically replicate data (using > hdfs, for auto. recovering ) ?, i had a hbase cluster on hdfs & after a > system crash, hbase cannot start, region servers cannot serve data , logs > showing that some blocks are missing , i had to wipe the data out & > recreate the tables > > i had always thought hbase which uses hdfs which replicates data can handle > such cases but that seems to be wrong > > thanx >
