Neo4j includes an effective tool for importing data into the Neo4j database. In version 4, it is called neo4j-admin import; in version 5, it is called neo4j-admin database import.

Neo4j 4: https://neo4j.com/docs/operations-manual/4.4/tools/neo4j-admin/neo4j-admin-import/

Neo4j 5: https://neo4j.com/docs/operations-manual/5/tutorial/neo4j-admin-import/
 

There are two steps often missed after the completion of the import. This article aims to address them.

Let's say you execute this import command in 4.4.x -

bin/neo4j-admin import --database=foo --nodes=import/movies.csv --nodes=import/actors.csv --relationships=import/roles.csv

After the completion of the import process, there will be two directories created:

<neo4j_home>/data/databases/foo

And

<neo4j_home>/data/transactions/foo

First thing to do:

Check the directory ownership for the above two directories. They should be owned by the user under which the neo4j service runs. If that is not the case. Execute the following command (using the Linux example and assuming the default Linux user/group being neo4j):

chown -R neo4j:neo4j <neo4j_home>/data/databases/foo

And

chown -R neo4j:neo4j <neo4j_home>/data/transactions/foo

Second, things are different in a cluster:

Before the database can be created, the newly created database directory, foo in the above example, must be present on all cluster members.

There are two ways of doing this. 

Option 1. Create an offline backup (dump) of the newly created database (created as a result of executing the import command) and load it on all other cluster members

bin/neo4j-admin dump --database=foo --to=/dumps/neo4j/foo.dump

bin/neo4j-admin load --from=/dumps/neo4j/foo.dump --database=foo --force

Next, create the database using cypher-shell or Neo4j Browser as:

CREATE DATABASE foo;

Option 2.  Create the database using the seed from the server where the import command was executed. 

CREATE DATABASE foo OPTIONS {existingData: ‘use’, existingDataSeedInstance: ‘<serverID>’}

This will copy the store files, for database foo in the above example, to all cluster members.

 

Note: The above points are valid for version 5 as well. However, the commands (in options1 and 2) may differ slightly.

 

Other Articles on the topic