Backup and Restore

There are two ways we backup the Neo4j graph:

• backing up the graph.db folder directory
• doing JSON dumps

Restoring graph.db folders is more convenient: you simply replace the current folder with the backed-up one. Restoring a JSON dump requires running the GraphSON command via the ./scripts/cmd wrapper, e.g:

./scripts/cmd graphson -d in /path/to/json/dump/file.json

This is best done on a dev machine rather than the server. Note that loading a JSON dump file should be loaded into a completely new database, i.e:

• copy the .json dump file to your local machine
• stop any running instances of Neo4j server (if any)
• set the $NEO4J_HOME variable to point to the dir of the Neo4j instance we’re loading into
• from the root of the ehri-rest project directory, run ./scripts/cmd -d in db_dump_201402030.json

Note: although the GraphSON loader does buffer input transactions (i.e. it doesn’t do the whole load in one TX) it has been observed to hit a java.lang.OutOfMemoryError: Java heap space error during this operation. If that occurs, increase the -Xmx1g arg in the ./scripts/cmd wrapper to, e.g., 2g.

Once loaded the Solr index should be fully re-built since Neo4j re-uses native graph ids and they will have changed.