Neo4j supports online backup and restores, or offline dump and loads (file system backups are not supported).
Online backup and restore is an Enterprise edition only feature, so is available locally when using Neo4j Desktop, but isn't available for our Kubernetes containers. The offline dump and load however, is the recommended way to copy data between environments.
On-the-fly disabling of replica's is coming soon.TM
Stopping and dumping of graphs in a Kubernetes cluster...
https://docs.human-connection.org/human-connection/deployment/volumes/neo4j-offline-backup
Backups consist of a backup of the Orchard Core SQL database, and also backups of the Published and Preview databases.
Restore the BACPAC file to a local SQL DB, or your own Azure SQL. You'll need to ensure that 'contained' database support is enabled for the SQL Server.
You can do this from management studio:
- Right-Click on the server instance, select Properties
- Select Advanced page, set under Containment the property value to True
We store content definitions in the SQL DB, so the restore should also give you a set of definitions.
Restoring dumps requires the use of the neo4j-admin.bat utility. It gets created in the /bin folder, when you create a graph in the Desktop edition of Neo4j. (Alternatively, it's installed when the Enterprise server edition is installed).
neo4j-admin.bat doesn't work out of the box, if you only have Neo4j Desktop installed. On Windows, it requires OracleJDK11 or ZuluJDK11. As ZuluJDK is open and free, this guide uses that.
(Todo: is this necessary if you use Open Terminal?)
Download and install the latest Windows msi from the download page. (Neo itself seems to use Zulu, can we just reuse its version?)
Chances are, you already have a version of a JDK installed, so you need to switch to the ZuluJDK. First check that your PATH environment variable has been updated with the Zulu bin before the existing JDK bin.
Next set the JAVA_HOME environment variable to the root folder for ZuluJDK, either with the SET command, or through the System Properties cpl, e.g.
set JAVA_HOME=I:\Program Files\Zulu\zulu-15\
Then to check the Zulu version is set correctly, run java -version, and check the output mentions Zulu, e.g.
openjdk version "15.0.1" 2020-10-20
OpenJDK Runtime Environment Zulu15.28+13-CA (build 15.0.1+8)
OpenJDK 64-Bit Server VM Zulu15.28+13-CA (build 15.0.1+8, mixed mode, sharing)
If the backup/dump is from an earlier version of Neo4j, click 'Manage', then 'Settings' and uncomment the following line and apply...
dbms.allow_upgrade=true
Unzip the appropriate backup.
Create a new graph through Desktop. Start it, click 'Open' to open the Browser, then create the preview graph...
CREATE DATABASE preview0
Stop the graph, then click 'Manage', then '>_ Open Terminal'
In the terminal, run these commands, replacing the path to the extracted backup (note the latest Neo db version allows spaces in the path, but earlier versions e.g. 4.0.4, don't)
cd bin
neo4j-admin.bat restore --from="I:\stax backups\neo4j-publish" --verbose --database=neo4j --force
neo4j-admin.bat restore --from="I:\stax backups\neo4j-preview" --verbose --database=preview0 --force
or if you're loading a dump...
cd bin
neo4j-admin.bat load --from="I:\stax backups\neo4j-publish" --database=neo4j --force
neo4j-admin.bat load --from="I:\stax backups\neo4j-publish" --database=preview0 --force
Install apoc.
Start the graph. The number of nodes and relationships should be in the millions.
Note that this article shows how to load a dump through the Desktop, but the options don't seem to be available.