thanks >The important content when a cluster is live is all in ZooKeeper.
do u mean if ZooKeeper is dead,the data will lose? in that case,ZooKeeper lost .META or .ROOT ,the data in hadoop will never be recover , thought there were some table folders in hadoop. 2010/6/18 Jean-Daniel Cryans <[email protected]> > Inline. > > J-D > On Thu, Jun 17, 2010 at 2:12 AM, 梁景明 <[email protected]> wrote: > > Something confused me for a long time. > > 1、how to recover hbase master? > > The master doesn't have anything to recover. You can start as many > masters as you want, one will be the active and the rest will just be > waiting for the active to die and take its place. The important > content when a cluster is live is all in ZooKeeper. > > > 2、how to recover hbase region server? > > The data is all in HDFS, you can lose up to 2 nodes at the same time > and yon won't lose data. Then its all up to Hadoop. > > > as hadoop ,there is checkpoint mirror to recover the hadoop master. > > As I said, there's nothing to backup for the master. > > > and now i only know to backup tables and restore them. > > To backup tables to can currently distcp the content of the hbase root > directory. See http://hadoop.apache.org/common/docs/r0.20.2/distcp.html. > There are other solutions, this mailing list is full of them. > > Also coming is https://issues.apache.org/jira/browse/HBASE-50 > > > one way to rescue hbase is to reinstall habse and create tables again . > > it cant write to shell or something ,all the stuff by manually. > > > > thanks for any help . > > >
