Question

Is there a way to split out a column family into multiple families - after the fact - in HBase?

This is clearly a bad idea, but in some cases it might be worthwhile, i.e.

1) if you have enough time to take your HBase server offline and see an opportunity to optimize some mapreduce jobs which are correlated to just a couple of columns.

2) you are importing from sqoop (no expert on sqoop but it appears by default to map each row of a database into a single column family - correct if mistaken).

3) you simply overlooked some access patterns early on in your app.

Related

Is there a way to consolidate column families into a new table (i.e. when cardinality deviates sufficiently that it would be better to do so for performance)?

Was it helpful?

Solution

We did it using Bulk load. We had to write MR job to split from one column family to 2 column families using HFileOutputFormat.

This output was loaded to hbase using completebulkload

Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top