How can I snapshot backup/restore JCR repository?
I need complete integrity. This means, single snapshot image produced by single backup command must be valid without any external dependancy. And it could be restored in other JCP instance completely. This means if I restore a snapshot image into a JCR instance, all existing data in the repository must be cleared and all data in the snapshot must be restored exactly as it was in previous repository.
It’s fine stopping whole services to do this. I prefer integrity at first.
How can I do this?
The JCR API itself doesn’t provide any easy way to do a complete and restorable backup of the entire repository (just like JDBC doesn’t cover backup/restore functionality of relational databases), so the answer to your question depends on which JCR implementation you’re using.
A simple solution that should works with any JCR implementation is to shut down the repository and do a filesystem / database backup of all the data used by the repository.
If you’re using Apache Jackrabbit, you can also use the backup feature included in the Jackrabbit standalone jar. Or if you prefer a Java API, you can access the same functionality using the RepositoryCopier class included in jackrabbit-core. When using the RepositoryCopier, you can even leave the source repository running as long as you can enforce on the application level that no client will attempt to write to the repository while the backup is running.