Thanks David, I will take a shot at this today.
Best Bhupesh On Sun, Aug 21, 2011 at 6:51 PM, David Burley <da...@geek.net> wrote: > Bhupesh, > > The following worked for us on a MySQL metastore: > > mysqldump metastore > metastore-backup-`date +%Y%m%d`.sql > mysql metastore > UPDATE SDS SET LOCATION=REPLACE(LOCATION, 'hdfs://hadoop-namenode-1', > 'hdfs://hadoop-namenode-2'); > UPDATE DBS SET `DESC`=REPLACE(`DESC`, 'hdfs://hadoop-namenode-1', > 'hdfs://hadoop-namenode-2'); > > But I think you noted you are using Derby. If I were using Derby, I'd > switch to MySQL to make the management less painful, and then do the > above. I provided some info on migrating from Derby to MySQL here: > > http://search-hadoop.com/m/J15J0FwAg1/v=plain > > If you must run Derby -- I'd think you could use the same tool, > RazorSQL to load up the metastore and then tweak the underlying data. > > Good Luck! > > --David > > On Fri, Aug 19, 2011 at 5:26 PM, Bhupesh Bansal <bhup...@groupon.com> > wrote: > > Hey Folks, > > > > I am wondering what is the easiest way to migrate data off one > hadoop/hive > > cluster to another. > > > > I distcp all data to new cluster, and then copied the metadata directory > to > > new cluster. > > hive comes up fine and show tables etc but the hive location is still > > pointing to old cluster > > > > There is one command > > alter table table_name set location new_location > > > > but it doesnt work for partitioned tables, is there a way we can do it > for > > *ALL* partitions easily ?? > > > > Best > > Bhupesh > > > > > > > ==== > This e- mail message is intended only for the named recipient(s) above. It > may contain confidential and privileged information. If you are not the > intended recipient you are hereby notified that any dissemination, > distribution or copying of this e-mail and any attachment(s) is strictly > prohibited. If you have received this e-mail in error, please immediately > notify the sender by replying to this e-mail and delete the message and any > attachment(s) from your system. Thank you. > >
