How should I copy a keyspace within a cluster

Question

I have a keyspace populated with data that was expensive to generate. I want two copies of this data within my cluster. I would like to end up with two keyspaces: lets call them mydata and mydatabackup, both of which contain identical data (I don't mind if the Cassandra timestamps are different).

Is there an easy way to do this? Closest thing I can find to an answer is to use sstable2json and json2sstable as suggested in response to a similar question? Is there a better way?

Answer 1

" Is there a better way?"

All Cassandra data are stored in the data/ folder (check config value data_file_directories in cassandra.yaml). You may also check the saved_caches_directory and commitlog_directory config.

Inside the data folder, you'll have

1. One folder per keyspace

2. One folder for system keyspace

3. Some folder for authentication etc..

   Inside each keyspace folder, you'll have

4. *-Data.db files which contain your real data

5. *-Filter.db files

6. *-Index.db files for index

7. ...

To replicate data, you do a plain copy of those folders.

In our team, the ops use a crontab to schedule regular backup of Cassandra data this way.

Note: sometimes, you may miss live data which are still in memory or in memtable and not flushed yet to disk. You can trigger a full compaction before backuping data files. But full compaction may hurt you perf so be careful

---

Better answer: use the provided tool to take a snapshot of you DB:

http://www.datastax.com/docs/1.0/operations/backup_restore