This content has been marked as final. Show 20 replies
user962305 wrote:No. The record is stored in a partition. A partition is a JE Database in a JE Environment. A Rep Group is a JE Replicated Environment. A Rep Group contains many partitions (JE Databases). A partition contains one or more key/value pairs.
When a record is stored on a node, is the partition also stored along with the record?
What I was trying to understand is how can data be redistributed if there is not link between record and partition stored some where?
Since the granularity for redistribution is partition, I agree we can identify the replication group (and therefore nodes in the replication group). But will we identify the data to redistribute?
We hash the key to get to a partition id. We then look up the partition id in a mapping table. If we migrate a partition, we change the mapping table. A key always maps to the same partition, no matter where it lives. The mapping table is persistent in the topology.
Edited by: Charles Lamb on Jan 7, 2012 10:00 AM
893771 wrote:A partition is a JE Database. We just look in the database to find all the records for that partition/database. Retrieving all records in a database is a simple matter using the JE API.
Let's say, we have partition 1, partition 3 and partition 7 on Server Node 1.
If we want to migrate partition 1, how will we know records that are in partition 1?
We know that those records are on Server Node 1 but how does Oracle NoSQL identify records related to a partition?