Hi Samir, We did this migration without any downtime. You can look at https://issues.apache.org/jira/browse/HBASE-12814.
Basically the steps are as follows: 1. Apply HBASE-12814 for 94 branch and 98 branches respectively. 2. Rolling restart all regionservers with the thrift replication parameter on. 3. Add a new replication source to have thrift protocol. 4. You can distcp (if you can afford downtime) or you can take a hbase snapshot and Export Snapshots if you can't 5. pause replication on the source side (no writes to destination) 6. upgrade hbase and hadoop 7. do the same on the other side 8. add a new peer for native replication (protobuf) make sure thrift backlog has drained 9. Revert HBASE-12814 patches I know its quite a few steps but we did this successfully with a few large clusters 1000+ nodes (upgrading from hadoop-1.x / hbase-94.x -> hadoop-2.x / hbase-98.x On Tue, Mar 1, 2016 at 3:33 AM, jaboy <[email protected]> wrote: > Hi Samir,Thanks for sharing the steps you performed. We are trying to > accomplish this migration on a production environment. Could you please > share the migration script template that worked in your cluster. Not sure > if > we need to include any special commands in the script other than just > 'distcp'. Thanks in advance. > > > > ----- > Thanks & Regards > Jaboy > -- > View this message in context: > http://apache-hbase.679495.n3.nabble.com/Hbase-data-migration-from-version-0-94-to-0-98-tp4078172p4078176.html > Sent from the HBase User mailing list archive at Nabble.com. >
