Mapr Vertica DR failover
This is what we need to achieved.
- 6 nodes production, 6 nodes DR
- Production: n001 …n006, DR d001…d006 (production and DR has different IPs)
- Vertica data “/vertica” hosted on mapr FS
- /vertica mirrored from production to DR
- Production Vertica is Up and Dr Vertica is standby (not running)
- In DR situation. Mirrored volume on DR site will be promoted from read-only to read/write
- Since IP is different on Production and DR, Vertica will not start up without modification
- Due to lacking of official procedure and best practice. I have come up with the follow procedure and tested it work.
Procedure to up an DR node from mirror data volume
- Edit admintool.conf on each DR notes (replace hostname and IP for new nodes on DR clusther)
- Edit config.cat for each database on each DR notes (replace hostname and IP for new nodes on DR clusther)
- Edit all files in Txnlog dir (replace hostname and IP for new nodes on DR clusther)
- Start vertica (notes, config.cat will be reverted back automatically on vertica startup)
- Down 1 node at a time (until all nodes are done)
- Run “alter node” to change node IP of the down node
- Restart the down node.
Will this be a recommended way to proceed?