There are various reasons you might want to migrate your storage node data - improving I/O performance, increasing capacity, or addressing failing hardware. Data can be migrated by adding/removing nodes, but often times it is simpler and more desirable to simply change move the rhq data directory. This can be accomplished, but doing so involves some manual steps which can be error prone.
What do people think about providing support for moving storage node data? I have attached a screenshot of the storage node configuration UI. We currently expose only some heap settings. We could add a section here for the data directory that would list the paths of the commit log, data, and saved caches directories. Changing the location of the data directory would result in the following,
* make sure that the target directory is writable * update cassandra.yaml with the new directory * restart the storage node * run repair as necessary
This encapsulates all of the work involved, making it easy for users to migrate data if/when necessary. Questions, comments, objections?
- John