How to implement centralized backups in Apache Cloudberry when gpbackup stores each segment’s data locally?

[tturgum5ekov]

Hello everyone 👋

I’m currently configuring backups for an Apache Cloudberry 2.0.0-incubating cluster using the gpbackup utility.
When I run the backup locally, everything works correctly — gpbackup creates files on the coordinator and on each segment host without errors.
However, I want all backups to be stored centrally on a separate backup server (connected via SSH).

Question:
Is there a simple or built-in way to:

• store all segment backups centrally on a single remote server,
• or use another tool that can write backups directly to remote storage (via SSH, NFS, or S3),
• so that gprestore can restore everything from one place?

Environment details:

• Apache Cloudberry: 2.0.0-incubating (based on PostgreSQL 14)
• Cluster setup: 1 coordinator + 2 segment hosts
• Backup tool: gpbackup v1.2.7-beta1+dev.7

Command used:
gpbackup --dbname postgres --backup-dir /backups/full --jobs 4 --leaf-partition-data --verbose

What I’ve tried:

• Using rsync to collect each segment’s /backups/full/seg*/... directories onto a single server.
• Writing a centralized backup script that connects via SSH to each node.
• Attempting to mount a shared NFS directory, but Cloudberry doesn’t automatically coordinate the mounts across all segment hosts.

[ostinru]

or use another tool that can write backups directly to remote storage (via SSH, NFS, or S3),

We are using wal-g to backup Greenplums and we adapted it to support Cloudberry as well. So, you may consider using it. Feel free to fill an issues on wal-g if you if you encounter problems or unclear documentation.

PS: we haven't tested wal-g with cloudberry PAX-storage yet. @leborchuk 's team is on the way to guarantee that PAX is backed-up properly.