Database backup

[carlonicora]

Hello!

First of all, thanks for your awesome work on the plugin.

I am exploring the possibility for the database backup, and it seems dozerDB does not support the online backup via neo4j-admin. Am I correct in thinking that you need to stop the database, run the backup, and then restart the database?

Thanks!

[gumshoes]

Hi, just a happy user here, what I do to backup and restore to a new DB when running Dozerdb in k8s/Docker is to stop the DB with stop database sourcedb and then exec into pod and:

export DBNAME=sourcedb
export NEWNAME=targetdb
neo4j-admin database info $DBNAME
time neo4j-admin database dump $DBNAME --to-path=/var/lib/neo4j/import/
cd /var/lib/neo4j/import
mv $DBNAME.dump $NEWNAME.dump
time neo4j-admin database load --verbose --from-path=/var/lib/neo4j/import $NEWNAME
chown -R neo4j:neo4j /data/databases/$NEWNAME

# Note that the load creates the files as root instead of neo4j which confuses neo4j and the only known fix is correcting the file perms and restarting the pod.  Can't switch to neo4j user and run the command unfortunately.

Back in Cypher:

# Start the source DB, create the new DB, and start it:
start database `sourcedb`
create database `targetdb`
# starting the new DB won't work via start cmd, you must restart the pod

:use `targetdb`

If there is a better way let me know, thanks.

[jmsuhy]

Your approach is a valid offline dump/load method and works as expected for Neo4j Community (and therefore DozerDB).

If you’re able to bring the service down, other commonly used options include stopping the database and taking a raw filesystem copy of the data and transactions directories of the DBMS, or using Kubernetes VolumeSnapshots.

Have you tried k8 VolumeSnapshots?

[gumshoes]

I use automated volume snapshots for scheduled backups but not for ad-hoc backups/clones, do you do that?

[carlonicora]

So far my best option is stop neo4j, run the backup and start it again, but it leaves me with a few minutes of downtime...

[olavgg]

I have good experience with ZFS snapshots. It works on a live database, but you may get a slow startup because it has to replay the transaction log. The safest option is to stop neo4j, snapshot, and start again.
Since you can send ZFS snapshot to another machine, you can do a normal offline dump on the remote machine.